TW548631B - System, method, and article of manufacture for a voice recognition system for identity authentication in order to gain access to data on the Internet - Google Patents

Abstract

A system, method and article of manufacture are provided for identifying a user through voice verification to allow the user to access data on a network. When the user requests access to data, such as a website, the user is prompted for a voice sample. Registration information about a user including a voice scan of the voice of the user is retrieved. The voice sample is received over the network and compared with the voice scan to verify an identity of the user. If the identity of the user is verified, data access is granted to the user. If the identity of the user is not verified, data access is denied.

Description

548631 A7 B7 V. Description of the Invention (1) Field of the Invention The present invention relates to a speech recognition system, especially a system that recognizes a user based on speech analysis to allow access to data on the network. BACKGROUND OF THE INVENTION There are many applications that often require users to access systems with high security requirements. Such applications include, but are not limited to, financial services such as confirmation and execution of stock transactions, bank account inquiries and wire transfers, Internet-based E-commerce, computer networks, safes, homes, doors, elevators, vehicles, and other high-priced equipment, etc., the scope of this description and the patent application will refer to these as "security systems." The physical token authentication devices currently used to identify individuals, such as secret cards or limited access cards, can be lost, stolen, lent to unauthorized individuals and / or copied, so security protection is somewhat inadequate. Another more sophisticated authentication method is called biometric authentication, which can provide higher security protection. The use of biometric authentication involves identification of unique physical characteristics, such as fingerprints, retinal scans, facial recognition, and speech pattern authentication. Please note that speech pattern authentication and speech pattern recognition are not the same in this paper and speech analysis techniques. In speech pattern recognition, the speaker first speaks a word (such as a word), and the system selects a pre-defined vocabulary to determine the spoken word. So speech recognition provides the ability to recognize the words spoken, not the ability to identify the speaker. The basis of retinal scanning is that each person's retinal blood vessel pattern is unique. 4Hickman200021tw; AND1P115.TW 1 This paper size applies to Chinese National Standard (CNS) A4 (210 X 297 mm) Page) Packing ---- Order --------- Line · Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7

V. Description of the invention (The pattern of the xJ is a fact that will not change in a lifetime. Although the retinal scan provides a lot of security, it is limited in practicality because it is expensive and requires complex hardware and software to implement it. Fingerprints and face recognition also require expensive and complex hardware and software to perform. Voice authentication, also known as voice authentication, voice mode authentication, speaker identification and voiceprint, can provide speaker identification functions. Voice identification and Voice authentication will be used interchangeably in this article, and both have the same meaning. The technology of voice authentication is widely discussed in the following U.S. patents: U.S. Patent No. 5,502,759; 5,499,288; 5,414,755; 5,365,574; 5,297,194; 5,216,720; 5,142,565; 5,127,043; 5,054,083; 5,023,901; 4,468,204, and 4,100,370, which are mentioned here for reference only. These patents describe many methods of speech authentication. Speech authentication is based on the speaker's pronunciation to identify the speaker completely. For example, using feature extraction and pattern matching algorithms can identify speaker hypotheses Identity, where the pattern comparison is a comparison of digitally received voiceprints with previously stored reference samples. Features used in speech processing include, for example, tone frequency, power spectrum chirp, spectral coefficients, and linear predictive coding, see B. S. Atal's (1976) Automatic recognition of speakers from their voice. Proc. IEEE, Vol · 64, ρ · 460-475, mentioned here for reference only. Other speech recognition technologies include, but are not limited to, Neural network processing, comparison of speech model with reference set, password authentication using selective adjustment of signal criticality, and simultaneous speech recognition and authentication. (Please read the note on the back? Matters before filling out this page) Printed by the Property Agency Staff Consumer Cooperative

4Hickman200021tw; AND1P115.TW 2

Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 V. Description of Invention (3) S. Furui's (1991) Speaker dependent—feature extraction, recognition and processing techniques. Speech communications, Vol. 10, pp · 505-520 There is a detailed description of the latest feature classification technology, which is mentioned here for reference only. Text-dependent speaker recognition must analyze predetermined pronunciations, while unrelated text recognition does not have to rely on any particular spoken text. However, in either case, the classifier generates a representative measure of the speaker and compares it with a preselected critical threshold. If the speaker's representative measurement is lower than the threshold, the identity of the speaker can be confirmed, otherwise the speaker is declared to be a fake person. The relatively low efficiency of speech recognition technology is one of the main reasons for its slow market entry. Equal Error Rate (EER) is a calculation method that involves two parameters: false acceptance (erroneous access permission) and false rejection (permitted access denial), both of which change according to the required level of secure access, However, as shown below, there is a trade-off between the two. The latest speech recognition algorithms (whether text-dependent or text-free) have an EER of about 2%. By changing the threshold of false rejection errors, false acceptance errors also change. The situation is shown in Figure 1 of J. Guavain, L. Lamel and B. Prouts (March, 1995) LIMSI 1995 scientific report. It is mentioned here for reference only. There are five graphs in the figure, which represent the relationships between the various false rejection rates (horizontal coordinates). The error acceptance rates of the speech authentication calculations show EER of 9.0%, 8.3%, 5.1%, 4.4%, and 3.5%, respectively. As mentioned earlier, there is a certain difference between the false rejection rate and the false acceptance rate. 4Hickman200021tw; AND1P115.TW 3 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) -------- --- Equipment -------- Order ---------. (Please read the notes on the back before filling out this page) 548631 A7 B7 V. Description of the invention (φ) A hyperbolic chart is drawn, where the lower EER 値 relationship chart is closer to the two axes. (Please read the note on the back? Matters before filling out this page) Therefore, if the system's false rejection rate is too low, the false acceptance rate will be too high, and vice versa. Various technologies for voice-based security systems are widely discussed in the following U.S. patents: U.S. Patent Nos. 5,265,191; 5,245,694; 4,864,642; 4,865,072; 4,821,027; 4,797,672; 4,590,604; 4,534,056; 4,020,285; 4,013,837; 3,991, No. 271; mentioned here for reference only. These patents describe the implementation of various voice security systems in different applications, such as telephone networks, computer networks, cars and elevators. However, none of these technologies can provide the sufficient performance required, because once the false rejection rate of these technologies is set too low, their high false acceptance rate will be unacceptable, and vice versa. According to the recommendation, the false rejection rate of speaker identification must be in the range of 1%, and the false acceptance rate must be in the range of 0.1% for the market to accept. Therefore, there is a wide need for improved false acceptance and rejection rates, and a reliable and secure voice authentication system would be of considerable benefit. Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economics Abstract of the Invention Provides a system, method, and manufactured product that allows users to access data on the Internet through voice authentication. When a user requests access to data, such as data on a website, the user is prompted to provide a voice sample. Then capture the user ’s 4Hickman200021tw; AND1P115.TW 4 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 548631 A7 B7 Printed clothing by the employee consumer cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs (Y) Registration information. The registration information includes a voice scan of the user's voice. Receive the user's voice sample through the network, and compare it with voice scanning to verify the user's identity. Once the user is identified, the user's data is granted access. If the user's authentication fails, they will be denied access to the data. In a specific embodiment of the present invention, the user's voice is recorded to establish a voice scan, and then the voice scan is stored. This can form part of the registration process. Voice scanning is best to include multiple groups of words spoken by the user to improve security, and identify the identity by comparing alternative words. For example, if the first word cannot confirm the identity of the user, the user is prompted to recite other words. Words. It is also possible to optionally record the time and date when the voice sample was taken together with the user's voice sample. BRIEF DESCRIPTION OF THE DRAWINGS The invention will be better understood with reference to the following detailed description. Each description refers to the attached drawings, wherein: FIG. 1 is a schematic diagram of the hardware implementation of a specific embodiment of the present invention; FIG. 2 illustrates a flowchart of a specific embodiment of the present invention, which uses speech analysis to detect emotions; 3 is a chart showing the average recognition accuracy of an S70 data set; FIG. 4 is a chart showing the average recognition accuracy of an S80 data set; FIG. 5 is a chart illustrating the average recognition accuracy of an S90 data set; Flow chart of a specific embodiment of the present invention for detecting emotions; 4Hickman200021tw; AND1P115.TW 5 This paper size applies to China National Standard (CNS) A4 specification (210 X 297 mm) ----------- «Packing -------- Order --------- Line · (Please read the notes on the back before filling out this page) 548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Description of the Invention (G) FIG. 7 is a flowchart of a method for detecting anxiety in a voice to prevent fraud in a business environment; FIG. 8 is a flowchart of a device for detecting emotion from a voice sample according to a specific embodiment of the present invention; FIG. 9 is a diagram showing a sound output from a sound according to a specific embodiment of the present invention. Flow chart of a device for recording images; FIG. 10 is a flowchart of a specific embodiment of the present invention that monitors emotions in a voice signal and provides feedback based on the detected emotions; FIG. 11 is a comparative use of specific embodiments of the present invention Flowchart of a person's speech signal and a computer-detected emotional speech signal to improve the emotion recognition of the invention, the user, or both; FIG. 12 is a block diagram of a speech recognition device made according to a specific embodiment of the present invention; Block diagram of element components and storage blocks; Figure 14 is a speech recognition system with a biological monitor and a pre-processor according to a specific embodiment of the present invention; Figure 15 is a biological signal generated by the biological monitor in Figure 14; Figure 16 is a biological Circuits in the monitor; Figure 17 is a block diagram of a pre-processor; Figure 18 illustrates the relationship between tone modification and biological signals; Figure 19 is a flowchart of a calibration routine; Figure 20 generally shows the improved tone achieved in the system of the present invention The structure of the candidate set during the period; Figure 21 is the use of voice recognition to identify users to allow users to access the network 4Hickman200021tw; AND1 P115.TW 6 This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm) ----------- installation -------- order ------ --- line (please read the notes on the back before filling this page) 548631 Printed clothing A7 B7 of the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the invention (/;) Flowchart of the specific embodiment of the present invention Figure 22 is a basic concept of a voice authentication system for controlling access to a security system; Figure 23 is a system for establishing a speaker identity according to the present invention; Figure 24 is a first step of an exemplary system for identifying a speaker according to the present invention; 25 is the second step of the system described in FIG. 24; FIG. 26 is the third step of the system described in FIG. 24; FIG. 27 is the fourth step of the speaker recognition system described in FIG. 24; A flowchart of a method for passing the border in a satisfactory manner; FIG. 29 is a method for identifying a speaker according to the present invention; FIG. 30 is another method for identifying a speaker according to the present invention; FIG. 31 is a diagram of a speaker identification system Basic components; Figure 32 is the information stored in Figure 31 speaker identification Examples in the information storage unit; FIG. 33 is a preferred embodiment of a speaker recognition system according to a specific embodiment of the present invention; FIG. 34 is more details of a specific embodiment of the speaker recognition system of FIG. 33; Flowchart of voice command recognition method for processing data on the Internet; Figure 36 is a general block diagram of an information system that uses voice signals to control content and applications on the network according to a specific embodiment of the present invention; Figures 37A, 37B, and 37C — 4Hickman200021tw; AND1P115.TW 7 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ------- ---- · Equipment -------- Order --------- (Please read the notes on the back before filling out this page) 548631 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of the Invention (g) FIG. 38 is a method of applying rules to form an acceptable sentence according to a specific embodiment of the present invention including a language translation function; and FIG. 39 is a representative hardware of a specific embodiment of the present invention including a language translation function Shi. DETAILED DESCRIPTION According to at least one specific embodiment of the present invention, a system is provided for performing various functions and activities through speech analysis and speech recognition. This system can be implemented based on a hardware implementation such as that shown in FIG. In addition, object-oriented programming (OOP) software programming can be used to implement various functionalities and user interface functions of the specific embodiments of the present invention. Hardware Description A representative hardware environment of a preferred embodiment of the present invention is shown in FIG. 1. This is a standard workstation hardware configuration with a central processing unit 110, such as a microprocessor, and a utilization system. The bus 112 interconnects many other units. The workstation shown in FIG. 1 includes a random access memory (RAM) 114, a read-only memory (ROM) 116, and an input / output adapter 118 that connects peripheral devices such as a disk storage device 120 to the bus Π2. , Connect other user interface devices such as keyboard 124, mouse 126, speaker 128, microphone 132, and / or touch screen (not shown) to the user interface adapter 122 of bus 112, connect the workstation to the communication network A communication adapter 134 (such as a data processing network) and a display adapter 136 connecting the bus 112 to the display device 138. 4Hickman200021tw: AND1P115.TW 8 is usually on the workstation. This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm). --------- (Please read the notes on the back before filling out this page) 548631 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of the invention (q) Install an operating system, such as Microsoft windows NT or Windows / 95 operating system (OS), IBM OS / 2 operating system, MAC OS or UNIX operating system. Software Description Object-oriented programming (OOP) is gaining popularity in the development of complex applications. As OOP gradually becomes the mainstream of software design and development, various software solutions must be adapted to use the advantages of OOP. To apply the principles of OOP to the messaging interface of an electronic messaging system, a set of OOP classes and objects must be provided for the messaging interface. OOP is a set of procedures for developing computer software using objects, including steps to analyze problems, design systems, and construct programs. Objects are software packages that contain data and a set of related structures and procedures. Because it contains data and a set of structures and procedures, it can be regarded as a self-contained component, and it can perform its specific tasks without the need for additional structures, procedures, or data. Therefore, OOP regards computer programs as a set of extremely independent components, called objects, each of which is responsible for a specific task. The concept of packaging data, structure, and programs in a component or module is called packaging. OOP components are generally software modules that can be used continuously. They provide an interface that conforms to a certain object model, and can be accessed through the component integration framework during execution. A component integration architecture is a set of architectural mechanisms that allow software modules in different processing spaces to utilize each other's capabilities or functions. Usually 4Hickman200021tw; AND1P115.TW 9 This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm) ----------- · Installation -------- Order-- ------- ^ _ vl (Please read the notes on the back before filling this page) 548631 A7 B7 V. Description of the invention σ) (Please read the notes on the back before filling this page) It is assuming a common component An object model on which to build a framework. It is necessary here to distinguish between objects and object categories. An object is a single instance of an object class. Object categories are often referred to directly as categories and can be viewed as a blueprint from which many objects can be formed. OOP allows programmers to create an object that is part of another object. For example, there is a composite relationship between an object representing a piston engine and an object representing a piston. In fact, a piston engine consists of a piston, a valve, and other components; the fact that a piston is an element in a piston engine can be represented logically and meaningfully in OOP using two objects. OOP printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs also allows the creation of new objects based on other objects. If there are two objects, one represents a piston engine, and the other represents a piston engine using a ceramic piston. A ceramic piston engine does not constitute a piston engine, but a piston engine with one more limitation than a simple piston; its piston is made of ceramic. At this time, the object representing the ceramic piston engine is called a derivative, inheriting all the layers representing the object of the piston engine, and adding more restrictions or details. The object representing the ceramic piston engine is "attached" to the object representing the piston engine. The relationship between these objects is called inheritance. When an object or category representing a ceramic piston engine inherits all levels of an object representing a piston engine, it also inherits the thermal characteristics of a standard piston defined in the piston engine category. But because the thermal characteristics of ceramic piston engines are usually the same as those of metal 4Hickman200021tw; AND1P115.TW 10 ^ The paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 548631 Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs B7 5. Description of the Invention (U) Pistons are different, so ceramic piston engines will replace the standard thermal characteristics with ceramic-specific thermal characteristics, abandon the original functions, and use new functions related to ceramic pistons. Different piston engines have different characteristics, but the related basic functions are the same (such as the number of engine pistons, ignition sequence, lubrication, etc.). When a programmer wants to access these functions in any piston engine object, they will call the same function under the same name, but each piston engine will have different / revised functions implemented behind the same name. The ability to hide the implementation of different functions under the same name is called a homogeneous vision, which can greatly simplify the communication between objects. Under the various concepts of compound relationship, packaging, inheritance, and homogenous vision, objects can represent almost everything in the real world. In fact, the only limitation on whether things can become objects in object-oriented software is our logical perception of the real world. Here are some standard categories: # Objects can represent physical objects, such as cars in traffic flow simulations, electrical components in circuit design programs, countries in economic models, or airplanes in air flow control systems. # Objects can represent elements in the computer-user environment, such as windows, menus, or graphical objects. # Objects can represent inventory, such as personnel files or tables of city latitude and longitude. # Objects can represent user-defined data types, such as time, angle, composite number, or point on a plane. 4Hickman200021tw; AND1P115.TW 11 This paper size applies to China National Standard (CNS) A4 specification (210 X 297 public love) ----------- · Installation -------- Order --- ------ Line · (Please read the notes on the back before filling out this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Description of Invention (\ >) Because OOP uses objects to represent logic The ability to separate things is so powerful that it allows software developers to design and implement computer programs that represent physical-level models, whether the physical thing is a physical entity, process, system, or combination of things. And because objects can represent everything, software developers can also create objects that will be used as components in larger software projects in the future. If 90% of the new OOP software programs are made up of existing components using existing reusable objects, then only 10% of the new software projects must be written and tested from scratch. Because 90% is taken from an extensively tested and reusable inventory of objects, there is only 10% chance of program errors. Therefore, OOP allows software personnel to create new objects using other objects that have already been constructed. This procedure is very similar to building a complex machine using components and subassemblies. OOP technology makes software engineering like hardware engineering. Developers can use various existing components of object types to build software. All this helps to improve software quality and speed up development. Various programming languages have begun to fully support OOP principles, such as packaging, inheritance, homogeneous vision, and compound relationships. The advent of the C ++ language has allowed many commercial software developers to embrace OOP. C ++ is an OOP language that provides fast code that can be executed by machines, and is suitable for commercial application and system programming projects. Now it has become a lot of OOP process 4Hickman200021tw; AND1P115.TW 12 This paper size is applicable to China National Standard (CNS) A4 specification (210 X 297 mm) -Order --------- line (Please read the note on the back? Matters before filling out this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs, Agricultural 548631 A7 V. Invention Description

The favorite choice of modern designers, but there are still many other OOP languages on the market, such as Smalltalk, Common Lisp Object System (CLOS) and Eiffel. In addition, many popular traditional computer programming languages, such as Pascal, have also added OOP capabilities. The advantages of object classes can be summarized as follows: • Objects and their corresponding classes break down complex programming problems into many smaller, simpler problems. • Packaging can organize the data into many independent small objects, and each small object can communicate with each other. This method is used to force the data to be streamlined. Packaging protects the data in the object from accidental damage, but allows other objects to call the constituent functions and structure of the object and interact with the data. • Use subclasses and inheritance to derive new types of objects from standard classes available in the system, and extend and modify objects. So to build new features, you don't have to start from scratch. • Homogeneous vision and multiple inheritance allow different programmers to mix and match many different categories of features, creating specialized objects that can be used in a desired way with related objects. • The class hierarchy and inclusion hierarchy provide a flexible mechanism for modeling relationships between real-world objects and objects. • Libraries that can use classes continuously are very useful in many cases, but there are some limitations. For example: • Complexity. The relevant class hierarchy in a complex system may contain dozens of 4Hickman200021tw: AND1P115.TW 13 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm). -------- Order ---- ----- Line (Please read the notes on the back before filling this page) 548631 A7 '^ ~ ------ B7 V. Description of the invention (丨 Lf〇 or hundreds of categories, it is easy to be confused. • Control flow. Programs written using class libraries are also responsible for controlling the flow (it must control the interaction of all objects created with a specific library). The programmer must decide when and what functions to call for what objects. • Double your efforts .Although category libraries allow programmers to use and reuse many small pieces of code, each programmer puts these pieces together in different ways. Two programmers can use the same set Class libraries write completely different internal structures (that is, designs), but use two programs with exactly the same purpose. The difference in the internal structure lies in the way the designer writes the program. It depends on many small decisions. These code snippets will inevitably perform the same work in slightly different ways. If put together, 'they should work well in theory, but the results are not ideal. The category library is very good. Flexibility. The more complex a program becomes, the more programmers have to re-create basic solutions to basic problems over and over again. A newer extension of the class library is the use of the class library framework. This framework is more For complexity, it consists of a large set of cooperative categories that complement small-scale models and main mechanisms that implement common needs and settings within a specific application. The original development was designed to allow application designers to display menus without having to worry about , Windows, dialog boxes, and the standard user interface of personal computers. 4Hickman200021tw; AND1P115.TW 14 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) (Please read the notes on the back before filling out (This page) | Install -------- Order ---------: Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 548631 B7____ 5. Description of the Invention (15) The framework also represents that the programmer's concept of interaction between the code written by him and the code written by others has changed. In the era of procedural programming, designers had to call The operating system provides a program to perform a specific task, but the program basically executes the entire code once from beginning to end, and the designer can only control the flow. If you only want to print paychecks, calculate mathematical forms or use only one The method execution method is appropriate to solve the problem. The development of graphical user interface has completely changed this procedural programming arrangement. The execution process is no longer controlled by program logic. Users can drive the program and decide when specific actions should be performed. Most personal computers today do this by monitoring the event loop of the mouse, keyboard, and other external event sources, calling specific parts of the programmer's code based on the actions performed by the user. Programmers no longer need to decide the order in which events occur. A program is divided into many small pieces that are separated and called in an unexpected order at an unexpected time. After developers transfer control to users, they write programs that are easier to use. However, the individual program fragments written by the developer still need to call the library provided by the operating system to complete the specific work, and the designer still has to decide the event loop to call the program flow control after each fragment program. Application code is still system-based. Even for event loop programs, programmers need to write a lot of code that does not have to be written separately for each application. The concept of the application framework is even more 4Hickman200021tw; AND1P115.TW 15 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) (Please read the note on the back before filling this page) -Install-- ------ Order --------- Line-Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Printed by the Employee Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 V. Description of the invention (丨 b) Further expansion Concept of the event loop. When using application frameworks, programmers no longer need to deal with building basic menus, windows, and dialog boxes, and then combine these details together. All basic user interface elements are ready, and designers can start processing applications directly. Code. Then, the general functions of the framework are replaced with the specific functions of the application to be designed, and construction is started. Application frameworks reduce the total number of program code that programmers must write from scratch. But because the frame is actually a general application that displays windows, supports copying and pasting, etc., designers can also give up more control than allowed by this event loop program. The framework code will pay attention to almost all event processing and control processes, and will only call the designer's code when the framework needs it (such as creating or processing proprietary data structures). A programmer writing a framework program can not only transfer control to the user (the same is true of the event loop), but can also transfer the detailed control flow in the program to the framework. Rather than separate programs, this method can use custom code to create more complex systems that work in the expected way, and iteratively builds for similar problems. As mentioned above, a framework is basically a set of fit categories for a reusable design solution for a hypothetical problem area. It usually contains objects (such as menus and windows) that provide default behavior, and programmers should inherit their parts when using

4Hickman200021tw; AND1P115.TW This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) -------------------- Order ------ --- Line (Please read the notes on the back before filling this page) 16 548631 A7 B7 V. Description of the invention (⑺) Pre-set behaviors, and then revise some other behaviors, so that the framework calls the application at the appropriate time. (Please read the notes on the back before filling out this page.) There are three major differences between the framework and the category library: • Differences in behavior and protocols. A category library is basically a collection of behaviors that are called when the program requires individual behaviors. And the framework not only provides behaviors, but also provides a protocol or a set of rules that control the combination of behaviors, including rules that regulate what programmers should provide and what the framework provides. • Differences between calling and revision. When using a class library, a program written by a designer creates objects and calls its member functions. You can also use the framework to create and call objects in the same way (that is, treat the framework as a class library), but if you want to make full use of the reusable design of the framework, the code written by the designer must distinguish the framework and be called by the framework. . The framework manages the control processes in its objects. Writing a program needs to distinguish the responsibilities of each piece of software called by the framework 'rather than specifying how different pieces of the program should fit together. Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs • Differences between implementation and design. Programmers can only implement reuse when using category libraries, but they can reuse designs when using frameworks. The framework contains a series of methods for working with related programs or software fragments. It represents a general design solution that can adapt to many specific problems within a known range. For example, a single framework can include the way the user interface works, but two different user interfaces created using the same framework may solve two distinct interface problems. 4Hickman200021tw; AND1P115.TW 17 This paper size applies to Chinese National Standard (CNS) A4 (210 X 297 mm) 548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of invention (\ each) A variety of problems and the framework required for programming work solutions can significantly reduce software design and development efforts. The preferred embodiment of the present invention is to implement a file using a hypertext markup language (HTML) in conjunction with a general-purpose secure communication protocol to transmit media between a user terminal on the Internet and a company. HTTP or other protocols can easily replace HTML without inappropriate situations. For information on these products, refer to "RFC 1866: Hypertext Markup Language-2.0" (Nov. 1995) by T. Berners-Lee, D. Connoly; and R. Fielding, H, Frystyk, T. Berners- "Hypertext Transfer Protocol-HTTP / 1.1: HTTP Working Group Internet Draft" by Lee, J. Gettys and JC Mogul (May 2, 1996). HTML is a simple data format used to create hypertext documents that can be carried across different platforms. HTML files are SGML files with general semantics that are suitable for expressing a wide range of information. World Wide Web's global information activities have been using HTML since 1990. HTML is an application of ISO standard 8879 '1986 Information Processing Text and Office Systems and Standard Generalized Markup Language (SGML). To date, the ability of web development tools to build dynamic web applications that connect clients with servers and interact with existing computing resources has its limits. Until recently, HTML was the main technology for developing web-based solutions. However, it has been confirmed that HTML is incompetent in the following areas: • Poor performance; 4Hickman200021tw; AND1P115.TW 18 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) --------- -Install -------- Order --------- line (please read the precautions on the back before filling this page) Explanation (q) • The user interface has limited functionality; • Can only generate static web pages; • Lacks the ability to interact with existing applications and data; and • Cannot be adjusted.

Sun Microsystem's Java language uses the following features to solve many client problems: • Improves client performance; • Allows users to create dynamic, real-time web applications; and • Provides the ability to create a wide range of user interface components. Developers can use Java to build robust user interface (UI) components. You can create custom "small devices" (such as real-time stock ticker, animated icons, etc.) and improve client performance. Java differs from HTML in that it supports the concept of client-side authentication and downloads appropriate processing to the client to improve performance. Can create dynamic real-time web pages. You can also use the custom UI components described above to create dynamic web pages.

Sun's Java language has emerged as the "Internet programming" language generally recognized by the industry. Sun defines Java as "simple, object-oriented, decentralized, interpretable, rugged, secure, free of architectural restrictions, portable, high-performance, multithreaded, dynamic, private, general-purpose programming language. The platform-less Java applet type supports Internet programming. " Java applet is Java 4Hickman200021tw; AND1P115.TW 19 which is compatible with Sun. This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm). ---- Order --------- ^ A_WT (Please read the notes on the back before filling out this page) A7 548631 B7_ 5. Description of the invention (丄 〇)

Application Programming Interface (API) are specialized small applications that allow developers to add "interactive content" to web documents (such as simple animations, web page decoration, basic games, etc.). The applet copies the server code to the client and executes it in a Java-compatible browser (such as Netscape Navigator). From a programming language perspective, Java's core functionality is based on C ++. Sim's Java documentation states that Java is basically "an extension of C ++ plus Objective C to provide a more dynamic approach to parsing." Another technology similar to Java is provided by Microsoft and ActiveX, which allows developers and web designers to create dynamic content on the Internet and personal computers. ActiveX includes tools for developing animation, 3D virtual reality, video and other multimedia content. These tools use Internet standards, can work on multiple platforms, and have been supported by more than 100 companies. The building block of a group, called ActiveX Control, is a small, fast component that enables developers to insert small pieces of software into Hypertext Markup Language (HTML) web pages. ActiveX Control can cooperate with many programming languages, including Microsoft Visual C ++, Borland Delphi, Microsoft Visual Basic programming system, and future Microsoft Java development tools, codenamed "Jakarta". ActiveX technology also includes ActiveX Server Framework, which allows development Personnel to create server applications. Skilled programmers can easily see that using ActiveX instead of Java can also implement the present invention without any discomfort. 4Hickman200021tw; AND1P115.TW 20 This paper applies Chinese National Standard (CNS) A4 Specifications (210 X 297 mm) (Please read the precautions on the back before filling out this page) Order --------- Information from the Intellectual Property Bureau of the Ministry of Economic AffairsΗConsumer Cooperative Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs Printed by the cooperative 548631 A7 B7 V. Description of the invention (> \) Emotion recognition The direction of the present invention is to apply the technology of identifying emotions in utterances to commercial use. Some specific embodiments of the present invention can be used to detect speech based on speech analysis The sentiment of the user, and then the detected emotion is output. His specific embodiment can be used to detect the emotional state in a call center conversation, and provide feedback to the switchboard or supervisor for monitoring. The specific embodiment of the present invention can also be used to sort the voice mail messages according to the emotion expressed by the caller. If already known For the target object, it is recommended to research several target objects to determine the most reliable part of the speech as an emotional indicator. If you do not know the target object, you can use other objects. The following discussion is in this direction: Professional actors get information because professional actors may over-emphasize specific discourse components and be prone to errors. • Data can be obtained from test subjects selected from a group that expects to be analyzed. This can improve accuracy. You can set the utterance of phone call quality (< 3.4 kHz) as the goal to improve the accuracy used in the telephone system. • Only rely on the voice signal for testing. That is to say, exclude new voice recognition technology because the new technology requires more Excellent signal quality and computing power. Data collection and evaluation test In the example, thirty people were recorded, and each person had to read four short sentences: 4Hickman200021tw; AND IP 115.TW 21 -----------, ΦΜ -------- Order --------- Line · (Please read the precautions on the back before filling in this page) This paper size is applicable to China National Standard (CNS) A4 (210 χ 297 mm)

548631 A7 B7 V. Description of the invention • “This is not what I expected.” • “I'll be right there.” (Please read the notes on the back before filling this page) • “Tomorrow is my birthday.” • “ I'm getting married next week. "Each sentence is recorded five times. During each recording, the subject should express one of the following emotional states: pleasant, angry, sad, scared / anxious, and normal (without emotion). Five subjects can also record sentences twice with different recording parameters. Therefore, each subject recorded a total of 20 or 40 utterances, for a total of 700 utterances, with 140 utterances for each emotional state. Each utterance can be recorded using a close-talking style (the distance between the microphone and the speaker's mouth is within 10 cm); the first 100 utterances are recorded at 22-kHz / 8bit, and the remaining 600 utterances are recorded at 22-kHz / 16bit Record. After the discourse set is established, an experiment can be performed to find the answers to the following questions: • How accurate are people without special training to express and recognize emotions in the discourse? • How accurately do people recognize their emotions in voices recorded 6 to 8 weeks ago? • Which emotions are the easiest / hardest to recognize? An important result of the trial of printing clothing for the consumer cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs was the selection of the most reliable set of discourses, that is, most people can recognize. This set of utterances can be used as training and test data for computer-implemented model recognition algorithms. 4Hickman200021tw; AND1P115.TW 22 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 548631 A7 B7 V. Description of the invention (above 3) Then use the existing interactive program to randomly select and play these words, Let users classify each utterance based on their emotional content. For example, twenty-three subjects can participate in the assessment, and twenty others participated in earlier recording trials. Table 1 is a confusion array of language usage based on the data obtained from the study of language usage. The rows and columns represent the real and judgment categories, respectively. For example, the second column shows that 11.9% of the words expressed with pleasant emotions were judged as normal (without emotions), 61.4% were judged correctly as pleasant, 10.1% were judged as angry, 4.1% were judged as sad, and 12.5% were judged to be afraid. It can also be seen from the table that the most recognizable category is anger (72.2%) and the most difficult to recognise category is fear (49.5%). There may be more confusion between sadness and fear, sadness and normality, and pleasure and fear. The average accuracy is 63.5%, which is similar to the results of other experimental studies. (Please read the precautions on the back before filling out this page) Packing -------- Order ----- S! Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs. Angry, sad and afraid total normal 66.3 2.5 7.0 18.2 6.0 100 Happy 11.9 61.4 10.1 4.1 12.5 100 Angry 10.6 5.2 72.2 5.6 6.3 100 Sad 11.8 1.0 4.7 68.3 14.3 100 Fear 11.8 9.4 5.1 24.2 49.5 100 4 Hickman200021tw; AND1P115.TW 23 This paper is for China National Standard (CNS) A4 specification (210 X 297 mm) 548631 A7

V. Description of the invention (above ...) Table 2 shows the statistics of the language used by the evaluator for each emotion category and the total performance of each category. It can be seen that the difference between anger and sadness is much lower than other emotion categories. Evaluator statistics, 'Table 2 Category average study bias Middle, minimum, maximum, normal 66.3 13.7 64.3 29.3 95.7 Pleasant 61.4 11.8 62.9 31.4 78.6 Angry 72.2 5.3 72.1 62.9 84.3 Sad 68.3 7.8 68.6 50.0 80.0 Fear 49.5 13.3 51.4 22.1 68.6 Total 317.7 28.9 314.3 253.6 355.7 (Please read the notes on the back before filling this page)

Table 3 below is the statistical performance of actors, that is, the accuracy of emotions expressed by subjects. More precisely, the numbers in the table indicate the degree to which other objects can correctly recognize the emotions expressed in a particular category. Comparing Table 2 and Table 3, we can find that the ability to express emotions (total average 値 62.9%) is approximately equal to the ability to identify emotions (total average 値 63.2%), but the expression bias in Table 3 is relatively large. ▽ Table 3 Statistics of actors (bookmarking line) m Printed category of the Intellectual Property Bureau of the Ministry of Economic Affairs Employee Consumer Cooperative Coordination Average Average Deviation Middle Minimum Minimum Maximum Normal 65.1 16.4 68.5 26.1 89.1 Happy 59.8 21.1 66.3 2.2 91.3 Angry 71.7 24.5 78.2 13.0 100.0 Grief 68.1 18.4 72.6 32.6 93.5 Fear 49.7 18.6 48.9 17.4 88.0 Total 314.3 52.5 315.2 213 445.7

4Hickman200021tw; AND1P115.TW 24 This paper size applies to Chinese National Standard (CNS) A4 (210 X 297 mm) A7 548631 B7 V. Description of the invention (25) (Please read the precautions on the back before filling this page) Table 4 It is self-reference statistics, that is, the accuracy with which an object recognizes its own words. It can be seen that people are more able to recognize their emotions (average 値 80.0%), especially angry (98.1%), sad (80.0%), and afraid (78.8%). Interestingly, identifying fear is more accurate than identifying pleasure. Some subjects are unable to recognize their happiness and normal mood. Table 4 Self-reference statistics: category average research deviation middle, minimum, maximum, normal 71.9 25.3 75.0 0.0 100.0 happy 71.2 33.0 75.0 0.0 100.0 angry 98.1 6.1 100.0 75.0 100.0 sad 80.0 22.0 81.2 25.0 100.0 afraid of 78.8 24.7 87.5 25.0 100.0 total 400.0 65.3 412.5 250.0 500.0 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs From the 700-sentence set, five objects (p = 70, 80, 90, 95, and 100%) containing at least 100% of households can be correctly identified A nested data set expressing the specified emotional discourse. In this paper, these data sets are represented by s70, s80, s90, s95, and slOO. Table 5 below is the number of elements in each data set. It can be seen that only 7.9% of the words were correctly recognized by all subjects. This increase linearly, reaching 52.7% by the time of the s70 dataset, which is equivalent to 70% of the interpretation of discourse emotions. 4Hickman200021tw; AND1P115.TW 25 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) 548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the invention (· ΐ4) V Table 5 P Level consistency data set Data set s70 s80 s90 s95 slOO Size 369 52.7% 257 36.7% 149 21.3% 94 13.4% 55 7.9% The above results provide valuable insights into the use of human language and can also be used as a benchmark for comparison with computer language use. Analysis of characteristics According to the findings, tone is the main voice prompt for emotion recognition. Strictly speaking, the tone is represented by the fundamental frequency (F0), which is the main (lowest) frequency of vocal cord vibration. Other sound variables that affect the emotional signal of a sound include: • Sound energy • Spectral characteristics • Formants (usually only the first one or two formants (F1, F2) are considered). • Time characteristics (speech rate and pauses). Another way to analyze the characteristics is to further strengthen the content of the feature set, considering some derived characteristics, such as the signal's LPC (linear predictive coding) parameters or the smoothest tone contour and its derivatives. The present invention can adopt the following strategies. First consider the fundamental frequency F0 (the main (minimum) frequency of vocal cord vibration), energy, speech rate, and the first three resonances 4Hickman200021tw; AND1P115.TW 26 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 Mm) -------------------- Order --------- line · (Please read the precautions on the back before filling this page) 548631 A7 B7 V. Description of the invention (&q; q) The peaks (FI, F2, and F3) and their bandwidths (BW1, BW2, and BW3) and the more the better the calculation. Then use the feature selection technique to arrange the statistics and select one of the most important features. Speaking rate can be calculated using the reciprocal of the average length of the vocal part of the discourse. For all other parameters, the following statistical 値 can be calculated: average 値, standard deviation 値, minimum 値, maximum 値, and range. In addition, the slope of F0 can also be calculated, which is the linear regression of the vocal part of the utterance, that is, the line most suitable for the contour of the tone. The relative sound energy can also be calculated by using the proportion of sound energy in the total speech energy. All of this adds up to about 40 characteristics in each discourse. Features can be selected using the RELIEF-F algorithm. For example, perform RELIEF-F for the s70 dataset, changing the most recent number from 1 to 12, and then rank the characteristics according to the sum of the grades. The first 14 characteristics obtained are as follows: F0 maximum 値, F0 standard deviation, F0 range, F0 average 値, BW1 average 値, BW2 average 値, energy standard deviation, speech rate, F0 slope, F1 maximum 値, energy maximum 値, energy Range, F2 range, and F1 range. Three sets of nested feature sets can be formed based on the sum of grades to investigate the influence of the features on the accuracy of emotion recognition. The first group includes the first eight characteristics (from F0 maximum 値 to the rate of speech), the second group is the first group followed by the next two characteristics (F0 slope and F1 maximum 値), and the third group contains all 14 characteristics . Regarding the calculation of RELIEF-F, "Estimating attributes: Analysis 4Hickman200021tw; AND1P115.TW 27" written by I. Kononenko in Proc. European Conf · On Machine Learning (1994) 27 This paper is based on the Chinese National Standard (CNS) A4 specification ( 210 x 297 mm) (Please read the notes on the back before filling out this page) ----- Order --------- Printed clothing by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Invention Detailed descriptions are provided in the "xg) and extension of RELIEF" pages 171-182, which are mentioned here for reference only. (Please read the notes on the back before filling out this page.) Figure 2 illustrates a specific embodiment of the present invention, which uses speech analysis to detect emotions. Job 200 receives a sound signal such as a microphone or a digitized sample. In operation 202, a predetermined number of voice signal characteristics are analyzed and selected according to the foregoing description. These characteristics include maximum fundamental frequency chirp, fundamental frequency standard deviation, fundamental frequency range, fundamental frequency average chirp, first formant average bandwidth, second formant average bandwidth, energy standard deviation, speech rate, fundamental frequency slope, first A formant maximum chirp, energy maximum chirp, energy range, second formant range, and first formant range. Using the characteristics selected in job 202, the relevant emotions in the speech signal are judged based on the extracted characteristics in job 204. Finally, the judged emotion is output at homework 206. For a description of judging emotions based on a voice signal according to the present invention, please refer to the discussion below, especially the reference discussions of FIGS. 8 and 9. The characteristics of the voice signal are preferably from the maximum fundamental frequency, the standard deviation of the fundamental frequency, the basic frequency range printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs, the average fundamental frequency, the average frequency of the first formant, and the average of the second formant. From the feature group consisting of bandwidth, energy standard deviation, and speech rate. Ideally, the extracted characteristics include at least one of the fundamental frequency slope and the largest chirp of the first formant. You can choose to extract some characteristics, including the fundamental frequency maximum 値, fundamental frequency standard deviation, fundamental frequency range, fundamental frequency average 値, average frequency of the first formant, second resonance 4Hickman200021tw; AND1P115.TW 28 This paper size applies to China Standard (CNS) A4 specifications (210 X 297 mm) 548631 A7 B7 5. Description of the invention (called) Peak average bandwidth, standard deviation of energy, and speech rate. The extracted characteristics preferably include the slope of the fundamental frequency and the maximum chirp of the first formant. Another optional feature includes the fundamental frequency maximum chirp, fundamental frequency standard deviation, fundamental frequency range, fundamental frequency average chirp, first formant average bandwidth, second formant average bandwidth, energy standard deviation, speech rate, Fundamental frequency slope, first formant maximum 値, energy maximum 値, energy range, second formant range, and first formant range. Computer Performance There are two non-standard ways to identify emotions in speech: a combination of neural networks and classifiers. The first method uses a two-layer rear-propagation neural network architecture with an input vector of 8, 10, or 14 elements, a hidden sigmoidal layer with 10 or 20 nodes, and an output linear layer with $ nodes. The number of outputs is the same as the number of emotion categories. Use s70, sso, and s90 to train and test calculus. These data sets can be randomly divided into training (67% of discourse) and testing (33%) subsets. Many neural network classifiers can be built using different initial weighting matrices. When applied to the s70 data set and the above 8 feature sets in this way, the average accuracy is about S5%, and the accuracy distribution of various emotion categories is as follows: normal state 40_50%, happy 55-60%, angry 60-80%, sadness 60-70%, and fear 20-40% ° 4Hickman200021tw; AND1P115.TW 29 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) -------- ---- · Clothing (Please read the precautions on the back before filling in this page)

Order --------- Line I Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economy 548631

5. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economics (30) The second method uses a combination of classifiers. Combining neural network classifiers with odd numbers, use initiation aggregation and cross-validation monitoring techniques to train with different training subsets. Portfolio decisions are made based on key selection principles. The recommended combination size is 7 to 15. Tune 3 is the average recognition accuracy for the s70 data set, all three sets of feature sets, and two neural network architectures (10 and 20 nerves in the hidden hierarchy, respectively). It can be seen that the pleasant accuracy of different feature sets and architectures is almost the same (~ 68%). Fear is less accurate (15-25%). The accuracy of anger was relatively low at 8 feature sets (40-45%), but it increased significantly by 14 feature sets (65%). However, the accuracy of sadness is higher in the eight feature sets than in the other sets. The average accuracy is about 55%. The fear of low accuracy is the same as the theoretical result. According to the theory, when the ratio of individual classifiers with unassociated errors exceeds 0.5 (0.6-0.8 in this example), the error rate of the selected combination increases. Figure 4 is the result of the s80 dataset. It can be seen that the accuracy of the normal state is very low (20-30%). Fear of accuracy varies widely, from 11% of 10 feature sets with 10 neural architectures to 53% of 10 feature sets with 10 neural architectures. Pleasure, anger, and sadness are relatively accurate (68-83%). The average accuracy (61%) is higher than the s70 data set. Figure 5 is the result of the s90 dataset. It can be seen that fear is more accurate (25-60%), but in the same form as the s80 dataset. Sad and angry quasi 4Hickman200021tw; AND1P115.TW 30 (Please read the precautions on the back before filling this page) China National Standard (CNS) A4 specification (210 X 297 mm)

Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 a7 B7 V. Invention Description G \) The accuracy is very high: anger is 75-100%, sadness is 88-93%. The average accuracy (62%) is approximately equal to the s80 dataset. Fig. 6 illustrates a specific embodiment of the present invention 'which uses statistics to detect emotions. First, a database is provided by job 600. The database contains speech parameter statistics related to human emotions, as shown in the above tables and Figures 3 to 5. In addition, the database contains many voice tones related to fear and many other tones related to pleasure, as well as the range of errors for specific tones. The next step is to receive voice signals during the job. Assignment 604 extracts one or more characteristics from the voice signal. For details on extracting features from a voice signal, see the section that discusses feature extraction earlier. Then in operation 606, the extracted speech characteristics are compared with the speech parameters in the database. Assignment 608 selects emotions from the database based on the comparison of the extracted speech characteristics and speech parameters. For example, compare the digitized utterance samples in the database with the digitized samples of the characteristics extracted from the speech signal to establish a possible emotional list, and then use calculations to consider statistics of human accuracy to identify emotions, and finally determine the most likely mood. The selected emotions are finally output at homework 610. Regarding the computerized mechanism for identifying emotions in speech, please refer to the "Demonstration Device for Detecting Emotions in Speech Signals" below. In one aspect of the invention, the database contains the possibility of specific speech characteristics related to emotions. The sentiment set selected from the database preferably includes analysis of the likelihood and selection of the most likely sentiment based on the likelihood. Database possibilities can optionally include language usage confusion statistics, such as the previous "Language Usage Mix 4Hickman200021tw; AND IP 115.TW 31 ----------- 装 -------- Order- -------- Line (Please read the precautions on the back before filling in this page) This paper size applies to China National Standard (CNS) A4 (210 x 297 mm)

548631 A7 B7 V. Description of Invention (3X) Confusion Matrix ". In addition, the statistical data in the database can optionally include automatic identification statistics as shown in the previous table. (Please read the precautions on the back before filling this page) In another aspect of the present invention, the extracted characteristics include the fundamental frequency maximum chirp, fundamental frequency standard deviation, fundamental frequency range, fundamental frequency average chirp, and the first formant Average frequency bandwidth, average frequency of second formants, standard deviation of energy, speech rate, fundamental frequency slope, maximum formant 値, maximum energy 値, energy range, second formant range and / or range. Figure 7 is a flowchart of a method for detecting anxiety in voice in a business environment to prevent fraud. The first is to receive a voice signal of a person in a business event at job 700. For example, pick up a microphone next to a person or use a voice signal captured by a telephone cable. Then in job 702, analyze the voice signals in the business event to determine the degree of anxiety of the personnel. The voice signal will be analyzed according to the previous instructions. Homework 704 will output an indication of the level of anxiety, preferably before the end of the business event, so that the scammer can evaluate whether he or she needs to be verified face-to-face before the person leaves. The output can take any form, including a printed print or a screen display on a computer screen. It must be understood that specific embodiments of the present invention can detect emotions other than anxiety, including stress and other emotions associated with deception. The specific embodiments of the present invention are particularly applicable to commercial areas such as contract negotiation, insurance negotiation, and customer service. These ranges of fraud can cost many companies millions of dollars a year. Fortunately, the present invention can provide assistance in eliminating fraud 4Hickman200021tw; AND1P115.TW 32 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 548631 A7 B7 V. The tool for explaining the invention (is). It should also be noted that the present invention is also applicable to the fields of law enforcement and court environment. It is best to output a certain degree of anxiety of the person to assist the fraudster in determining whether the words spoken by the person are deceptive. This can be determined based on the statistics of the specific embodiment of the present invention discussed above with reference to FIG. You can choose to output the speaker's anxiety indicator in real time, so that the fraud prevention person can get the result quickly, and ask the questioner immediately after the other person speaks suspicious words. Alternatively, an alert may be included in the anxiety indicator, and an alert may be issued when the anxiety exceeds a predetermined level. Alerts can include visual notifications or siren sounds on a computer screen to alert supervisors, listeners, and / or detect fraud. You can also connect an alarm to a recording device and start monitoring the conversation as soon as the alarm is issued without the conversation being monitored. The alert option is ideal for situations where many people take turns speaking. For example, the customer service department or a call to a customer service representative. As each customer takes turns speaking to a customer service representative, the present invention will detect the degree of anxiety in the customer's words. If the customer's anxiety level exceeds a predetermined level and an alarm is triggered, a visual indication on the customer service representative's computer screen or a flashing light can notify the service representative. The customer service representative will then know that there may be deceptive behavior and wait for the opportunity to expose it after it is determined. Alerts can also be used to notify supervisors. Alternatively, you can start monitoring the conversation after the alarm is triggered.

4Hickman200021tw; AND1P115.TW (Please read the notes on the back before filling this page)

ϋ ϋ · I ϋ ϋ · ϋ ϋ 一 -ον · I ϋ ϋ I n ϋ ϋ I

Printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs This paper is sized to the Chinese National Standard (CNS) A4 (210 X 297 mm)

A7 548631 B7 _ 5. Description of the invention (3 $) In a specific embodiment of the present invention, at least one characteristic of the voice signal must be extracted and used to judge the speaker's anxiety. Possible characteristics include the fundamental frequency maximum chirp, fundamental frequency standard deviation, fundamental frequency range, fundamental frequency average chirp, first formant average bandwidth, second formant average bandwidth, energy standard deviation, speech rate, fundamental frequency The slope, the maximum formant of the first formant, the maximum form of energy, the energy range, the second formant range, and the first formant range. So, for example, the degree of tonal sway judged from the fundamental frequency reading can be used to help determine the level of anxiety. The greater the swing, the higher the level of anxiety. It is also possible to consider pauses in the speaker's discourse. The next section describes devices that can be used to judge emotions, including anxiety, in a voice signal. Exemplary Devices for Detecting Emotions in Speech Signals This section discusses several devices for analyzing speech in accordance with the present invention. A specific embodiment of the present invention includes a device that analyzes the speaker's utterances to determine the state of emotion and thread. The analyzer analyzes the instantaneous frequency or tonal components of the ~ formant frequency band in human speech. When analyzing the utterance, the device analyzes specific radon occurrence patterns such as the differential first formant tone, tone change rate, duration, and time distribution pattern. The relationship between these coefficients and instant and long-term emotions and states is very complicated, and they are all basic coefficients of emotion.

4Hickman200021tw; AND1P115.TW This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm) ------------ Installation ---- (Please read the precautions on the back first (Fill in this page) Order --------- line t Printed by the Employees 'Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Printed by 548631 A7 B7 V. Invention Description) Human speaking is made by two The basic vocal mechanism begins. Vocal cords: Thin, stretchy films controlled by muscles that vibrate as air from the lungs passes through them. The “fundamental sound” of “quoting buzz” is between 80Hz and 240Hz. Intentional and unintentional muscle contraction and relaxation will cause this frequency to change within a proper range. The basic waveform of `` buzz '' contains many harmonic Waves, some of which can stimulate many fixed and variable cavity resonances related to the channel. The second basic sound produced by speaking is a pseudo-random noise with a very wide and consistent frequency distribution. This is the chaos formed by the air sent out through the channel. Caused by the flow is called "hisse." Most random murmurs are adjusted by tongue movement and also stimulate fixed and variable cavities. The words spoken are a complex mixture of "hum" and "hiss", shaped and pronounced in a resonance chamber. When analyzing the energy distribution in speech, it can be found that the energy is divided into several distinct frequency bands, called formants. There are three main formants. The system discussed here makes use of a first formant frequency band extending from a basic "fizz" frequency to about 1000 Hz. This formant lacks the highest energy, and it also reflects the high frequency modulation caused by various channels and facial muscle tension changes. In fact, analyzing specific first formant frequency distribution patterns can measure discourse-related muscle tension changes and interactions. Since these muscles are mainly deflected and vocalized by the second unconscious program, while the unconscious program is affected by the state of emotion, we can judge the relative measurement of emotional activity whether or not the speaker notices this state. The research results also support general assumptions, because the speaking mechanism is very complex and highly independent. 4Hickman200021tw; AND1P115.TW 35 (Please read the precautions on the back before filling this page) Order --------- line i This paper size applies to China National Standard (CNS) A4 (210 X 297 mm)

548631 A7

V. Description of the invention (Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Winter Economics, so few people can consciously "project" fictional emotional states. In fact, "want to express fictional emotions, they often produce themselves in voice graphics Unique psychological pressure "fingerprint". Due to the characteristic of the first formant speech sound, the present invention analyzes the first formant speech signal of FM demodulation and generates an output indication of zero chirp. The zero chirp or " The frequency or number of "flat" points, the length of the zeros, and the ratio of the total time of the zeros to the total time of the singles are all indicators of the emotional state of the person. As long as the user looks at the output of the device, he can see or feel zeros. The presence of 値, and then observing the number or frequency output of zero 値, the length of zero 値, and the ratio of the total time of zero 値 existence to the length of the single word period in the word period can determine the emotional state of the speaker. The first formant frequency band is the FM demodulation signal, which is applied to the single-word detection circuit that detects the presence of the FM demodulation signal. FM The demodulated signal is also applied to the zero-line detection tool, which detects the parts in the FM demodulated signal and generates an output indication. The output circuit is then connected to the word detector and zero-line detector. Word detection When the device detects the presence of the FM demodulated signal, it will start the output circuit, and the output circuit will generate an output indication with or without zero in the FM demodulated signal. The output display mode of the output circuit allows the user to perceive the output. The user provides an indication of the presence of zero noise in the FM demodulation signal. As long as the user of the device monitors the zero noise, he can judge the emotional state of the voice analysis object. Specifications (210 X 297 mm) ----------- • installation -------- order --------- line · (Please read the precautions on the back first (Fill in this page) Printed clothing by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Description of the invention (In another specific embodiment of the present invention, the head is analyzed. The so-called vibrato is generated by semi-random response, Cooperating with the following other reactions is to investigate fraud Valuable data: breathing volume; inspiration-expiration ratio; metabolism ratio · 'Respiration regularity and ratio; words and concepts; facial expressions · motorogen response; and some specific drug responses. But' previously尙 No technology has been developed that can effectively and reliably analyze sound changes, and clinically judge the subject's emotional state, opinions, or attempt to deceive. Early experiments that tried to associate changes in sound quality with emotional stimuli confirmed that human speech is greatly affected by emotions. The detectable changes in speech measured by stress stimuli are much faster than the physiological signs indicated by the traditional autonomic nervous system. Stress causes two kinds of sound changes. The first is only caused by extreme pressure These changes are manifested in perceptible changes such as speech rate, volume, tremor of the sound, changes in distance between syllables, and changes in the basic pitch or frequency of the sound. When the level of stress is below the level of total runaway, at least some subjects will consciously control this total change. The second change in sound is a change in sound quality. The human ear cannot recognize such changes, but it is clear that the vocal cords are slightly tightened under slight pressure, which results in a subconscious expression that suppresses the selected frequency change. When expressed graphically, it is easy to recognize pronunciations with no or normal vocalization and moderate stress, attempts to deceive, or hostility. These models are of both sexes, all ages, and 4Hickman200021tw; AND1P115.TW 37 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ------------ installation— ( Please read the precautions on the back before filling this page) Order --------- Line 0 · 548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the Invention (3 teeth) Exactly. The second change is not controlled by consciousness. The human vocal structure makes two sounds. The first sound is caused by vocal cord vibration, and the vocal cord vibration is caused when the glottis is partially closed, and the lung cavity and lung contraction force air through the glottis. Depending on the gender and age of the speaker, and the tone used by the speaker, the frequency of these vibrations typically varies between 100 and 300 Hz. This sound decays quickly. The second kind of sound involves the formant frequency, which constitutes the sound caused by the resonance of the various chambers of the head, including the larynx, oral cavity, nasal cavity, and sinus cavity. In the case of vocal cords, ‘This sound is caused by a lower-frequency sound source stimulating the resonance chamber, and in the case of silent friction, it is caused by the passage of air from the lungs being partially restricted. Regardless of the stimulus source, the frequency of the formant is determined by the resonance frequency of the relevant chamber. The formant frequency is usually around 800 Hz and appears in a well-defined frequency band consistent with the resonance frequency of each chamber. The first (or lowest) formant is the formant caused by the oral cavity and the larynx. It should be noted that its frequency will move with the change in size and capacity of the oral cavity when forming various sounds, especially when the vowel is pronounced . The highest formant frequency is relatively stable because the volume of the chamber is relatively stable. The formant waveform is different from the rapidly decaying vocal cord signal, which is a continuous signal. When a sound is emitted, the sound waveform is loaded onto the formant waveform by means of amplitude modulation. It was found that a third type of signal exists in human sounds, and that this type of signal is related to the second type of sound change discussed earlier. This type of signal is subsonic, 4Hickman200021tw; AND1P115.TW 38 (Please read the precautions on the back before this page)

-, 1T This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) 548631 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 __ B7________ Description of the invention (3 ^) This is the frequency tone that human hearing cannot hear Change, vocal cord sound and formant sound are more or less present. This signal is usually between 8 and 12 Hz, so the human ear cannot hear it. Because this characteristic constitutes frequency modulation, not amplitude modulation, it cannot be detected directly in the time base / amplitude map record. Because this sub-sound signal is one of the important sound indicators of psychological stress, it needs to be processed more deeply. There are many analogs that can be used to provide a graphical representation of the entire sounding program. And mechanical and electronic analogs have been adopted smoothly, such as designing computer voice. However, these analogs both see the acoustic signal source (vocal cord) and the cavity wall as hard and constant properties. However, the vocal cords and the main cavity walls that form the resonance peaks are actually elastic cellular tissues, which sensitively respond to all movements of the complex muscle groups that control the cellular tissues. These muscles that control the vocal cords using bones and connections allow people to intentionally and automatically make sounds and tonal vibrations. Similarly, controlling the muscles of the tongue, lips, and throat allows intentional and automatic control of the frequency of the first formant. The other formants are also affected, but to a lesser extent. What I noticed was that during normal speech, these muscle movements are only a small part of their total working ability. Therefore, even if these muscles are used to change the position of the vocal cords and the position of the lips, tongue, and inner throat wall, the muscles remain relatively relaxed. Studies have shown that in this relatively relaxed state, natural muscle ups and downs usually occur at the 8-12 Hz frequency mentioned earlier. This undulation causes a slight change in the vocal cord tension and also changes the fundamental tone frequency of the sound. 4Hickman200021tw; AND1P115.TW 39 This paper is again applicable to the Chinese National Standard (CNS) A4 (210 X 297 mm) 548631 A7 B7 Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economy The capacity of the resonance cavity (especially the resonance cavity related to the first formant) and the elasticity of the cavity wall are also slightly changed, so that the formant frequency is changed. These changes in center frequency constitute a modulation of the center or carrier frequency. It is important to note that no matter what the fundamental tone frequency or formant frequency of the sound is, the listener cannot detect it, partly because the change is small, and partly because its frequency is mainly at the aforementioned unseen frequency Within range. In order to observe this frequency modulation, any of the many existing frequency modulation techniques can be used. Of course, it is important to remember that the modulation frequency is between 8-12 Hz, and the carrier is a frequency band in the sound spectrum. In order to fully understand the above description, we must first understand the concept of "mass center" of this waveform. We can judge the center point between the two ends of any stroke of the recorder. If the center points between the two ends of all the strokes are marked, and then the center points are connected into a continuous curve, it can be seen that a line is approximately equal to the average 値 or “center of mass” of the entire waveform. Connect these marks together and make a smooth correction to produce a smooth curve. This line represents the subsonic frequency modulation caused by the previously mentioned undulations. As mentioned above, the muscle groups associated with the vocal cords and the cavity wall are subject to mild muscle tone when mild or moderate psychological stress is applied in individual experiments. Subject cannot detect this tension, even the inspector, without assisted observation technology 4Hickman200021tw; AND1P115.TW 40 This paper size applies Chinese National Standard (CNS) A4 (210X297 mm) (Please read the precautions on the back first (Continued on this page) — installed. This-line 548631 A7 B7 printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. The invention description (M) cannot be noticed. However, this tension is sufficient to reduce or almost eliminate muscle undulations in an unstressed subject, thus eliminating the basis for carrier frequency changes that produce subsonic frequency modulation. Although only the technology that uses sound as a physiological medium for psychological stress assessment uses subsonic waveforms, but in the currently used technologies and devices that can detect physiological changes, sound does not provide inaudible psychology caused by psychological stress. Variations of additional instrument measurements. Among the four commonly-used physiological changes mentioned above (electroencephalogram, cardiac activity, skin conductivity, and respiratory activity), respiratory activity and cardiac activity directly and indirectly affect the amplitude and details of the vocal waveform and provide more overall psychology The basis of stress assessment is more apparent in tests involving continuous acoustic response. Figure 8 shows another device. As shown in the figure, the converter 800 converts an object's spoken sound wave into an electrical signal waveform and connects it to the input terminal of the sound amplifier 802. This amplifier just amplifies the power of electrical signals to a more stable and practical level. The output of the amplifier 802 is connected to a filter 804. This filter is mainly to eliminate some unwanted low frequency and noise components. After the signal is filtered, it is connected to the FM frequency discriminator 806 to convert the deviation from the center frequency into a signal that varies in amplitude. The amplitude-changing signal is then detected by the detection circuit 808 to rectify the signal to generate a signal composed of a series of half-wave pulses. After detection, connect the signal to the integration circuit 810 to integrate the signal to the required level. Circuit 810 will integrate the signal to a small range 4Hickman200021tw; AND1P115.TW 41 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (2 丨 〇 > < 297 mm) 548631 Five ______ Intellectual Property Bureau of the Ministry of Economic Affairs Printed by the employee consumer cooperative A7 B7 Description of the invention (φι) Set to 1 to generate a waveform, or integrate to a greater degree to generate a signal. After the signal is integrated, it is amplified by the amplifier 812 and then connected to the processor 814 to judge the emotions related to the sound signal. Finally, an output device 816 such as a computer screen or a printer is used to output the detected emotion. Statistics can also be selectively output. FIG. 9 shows a simpler device for generating a visible record according to a specific embodiment of the present invention. A sound signal is converted to an electrical signal by a microphone 900 and recorded to a tape device 902 by magnetic action. Then, the signal is processed by the back device at various speeds at any time, and the played signal is connected to the conventional diode 904 for rectification. The rectified signal is connected to the movable contact of the conventional selector 906 and 908 selector switch. The movable contact of the switch 908 can be switched to any fixed contact, and a capacitor is connected to each contact. Figure 9 shows a group of four capacitors 910, 912, 914, and 916, one of which is connected to the fixed contact of the switch and the other is grounded. The output of the amplifier 906 is connected to the processor 918. The recorder used in this particular combination of devices is a Uher 4000 four-speed recording device with a built-in amplifier. Capacitors 910-916 have 0.5, 3, 10, and 50 microfarads (ixF), respectively, and the input impedance of amplifier 906 is about 10,000 ohms (10 kQ). As we will see later, this device can be, or other various components have been used. In the operation of the circuit in Figure 9, the waveform rectified by the diode 904 will be integrated to 4Hickman200021tw; AND1P115.TW 42 (Please read the note on the back first; this page). The size of this paper applies to the Chinese National Standard (CNS) A4 specification. (210x297 mm)

548631 A7 B7 V. Description of the invention (Battle 3) The degree of time required is selected. The time constant is selected so that the frequency-modulated subsonic result is displayed as a DC level that changes slowly along the line representing the "center of mass" of the waveform. The travel in this particular diagram is relatively fast, indicating that the switch is connected to a lower capacitor. In this specific embodiment, the composite filtering is performed by the capacitor 910, 912, 914, or 916, and is performed by the recorder during deceleration playback. Telephone Job with Switchboard Feedback FIG. 10 illustrates a specific embodiment of the present invention that monitors emotions in a voice signal and provides switchboard feedback based on the detected emotions. The first is to receive a voice signal representative of the conversation component between at least two subjects at home 1000. Then, the emotion related to the voice signal is determined in operation 1002. Finally, in the assignment 1004, a feedback is provided to a third party according to the judged emotion. The conversation can take place over a telecommunications network or, if used with an Internet phone, over a wide area network. You can choose to filter your emotions. You only need to provide feedback if you judge that the emotions belong to the negative emotions selected from the negative emotion group consisting of anger, sadness, and fear. The same setting can be made based on positive or neutral emotions. The emotion can be judged by extracting characteristics from the voice signal according to the previous detailed description. The invention is particularly suitable for operations related to emergency response systems, such as the 911 system. The invention can be used in such systems to monitor incoming calls. Call answering technology 4Hickman200021tw; AND1P115.TW 43 This paper size is applicable to China National Standard (CNS) Α4 size (210X297 mm) (Please read the precautions on the back before writing this page)

Ml., Τ Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of Invention (w) A technician can judge the emotion of the caller when talking with the caller. The emotions are then transmitted via radio waves to the emergency response team, the police 'fire and / or ambulance personnel, to make them aware of the emotional state of the caller. Another scenario is that one of the objects is a customer, the other is an employee ’such as an employee employed by a call center or customer service department, and a third party is the supervisor. In this example, the present invention can be used to monitor conversations between customers and employees to determine whether the customers and / or employees are impatient. Once negative emotions are detected, they are sent back to the supervisor, who then evaluates the situation ’and intervenes if necessary. Improve emotion recognition FIG. 11 illustrates a specific embodiment of the present invention, which compares the user and computer voice signal emotion detection to improve the emotion recognition of the invention, the user, or both. ° First, the voice signal and the voice signal are provided in operation 1100. Related emotions. Assignment 1102 will automatically determine the emotions associated with the voice signal using the methods previously described. Automatically judged emotions are stored ' in homework 1104, for example, on a computer-readable medium. Assignment 1106 receives user judgment emotions related to the voice signal determined by the user. Assignment 110: Stomach Compare the emotion judged automatically by the user. The invention can send or receive voice signals. Optionally confirm the relevant emotions in the voice signal based on the emotions provided. At this time, it should be judged whether the emotion judged automatically or the emotion judged by the user is consistent with the confirmed emotion. If used

4Hickman200021tw; AND1P115.TW 44 Please read the notes on the back first

The paper size of the edition applies to the Chinese National Standard (CNS) A4 specification (2Ι 10'〆297 mm) 548631 A7 B7 V. Description of the invention U5) (Please read the notes on the back before filling this page) Matches the mood, this user should win the prize. In addition, you can extract at least one of the characteristics of the voice signal and automatically determine the emotion in it, as in the case described earlier. To assist users in identifying emotions, an emotion recognition game may be played according to a specific embodiment of the present invention. Games allow users to compete with computers or other people to see who can best identify emotions in conversation recordings. One of the practical applications of the game is to help lonely people develop better emotional skills by identifying emotions in conversation. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs According to a specific embodiment of the present invention, a device can be used to create voice signal related data that can be used to improve emotion recognition. In such a specific embodiment, the device accepts voice through a microphone or a voice recorder. After the actual sound waves are converted into electrical signals, they are transmitted in parallel to a set of standard filters covering the sound frequency range. Setting the center frequency of the lowest sprinkler to any chirp that allows the electrical energy of the voice signal amplitude containing the lowest sound frequency signal to pass through will create a subsequent filter until the last transmitted energy, usually between 8kHz to 16kHz or 10kHz Between 20kHz, the center of all filters is chirped and determines the exact number of filters. The center frequency of the first filter does not matter, as long as it can capture the lowest tone of human voice, about 70Hz. Basically any commercially available filter bank can be used as long as it can be matched with a commercially available digital tablet and microcomputer. The paragraphs of the description describe specific center frequencies and microprocessor sets in the preferred embodiment. The enhanced calculations disclosed in this manual will also filter 4Hickman200021tw; AND1P115.TW 45 This paper size applies Chinese National Standard (CNS) A4 specifications (210X297 mm) '548631 A7 B7 V. Description of the invention (recognition) ^ ^ The average quality is replaced by an acceptable frequency and amplitude, so the quality of the filter is not particularly important. Of course, after calculating the center frequency, the bandwidth of all filters is defined by a ratio of 1/3. After the filter is divided, a commercially available digitizer or multiplexer and digitizer are used to digitize the output voltage of the filter. In the case of the disclosed preferred embodiment, the digitizer and the confirmed Commercial filters are assembled together to eliminate connection circuitry and hardware. The quality of the conversion or recognition speed of the digitizer is not very important, because the correction calculation (reference specification) and sampling rate requirements of this application are very low. The average quality of digitizers currently on the market is higher than this requirement. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Any complex sound that carries changing information can be reduced to a similar value by capturing the peak frequency and amplitude of the signal to reduce the number of bits of information. Of course, it is old knowledge to perform such operations on discourse signals. However, in the study of discourse, many specific regions where such peaks occur are often classified as "formant" regions. However, these area approximations are not necessarily consistent with the peaks of each speaker in all circumstances. Discourse researchers and previous inventions have focused on measuring peaks that fall within the frequency range of standard formants and calling them "correct" peaks. It seems that there is no estimate in their definitions, only absolutes. This phenomenon has led many research and formant measurement devices to manually eliminate many of the appropriate peaks required to instantly and appropriately represent complex and widely varying sound waves. Since this discussion is intended to be applicable to animal sounds and all human languages, it is not intended to discuss formants 4Hickman200021tw; AND1P115.TW 46 This paper size applies Chinese National Standard (CNS) A4 specifications (210X297 mm) 548631 A7 A7 B7 Economy Printed by the Ministry of Intellectual Property Bureau's Consumer Cooperative Cooperative, V. Invention Description (w) and other artificial restrictions, we will treat sound waves as sound waves that can analyze the complexity and high variation of any sound. In order to standardize and simplify the peak chirp identification operation, regardless of filter bandwidth, quality, and changes in digitizer discrimination, the actual stored amplitudes and frequencies 値 are "representative 値". Therefore, the digital representation of the bandwidth of the high-frequency filter is the same as the bandwidth of the low-frequency filter. Each filter is assigned a continuous 値 from 1 to 25, and the sound size is distinguished by a number from 1 to 40 for display on the CRT screen. If the output amplitude of the filter on the right of the peak chirp filter is greater than the output of the filter on the left, as long as the number of the filter is increased close to the decimal 値 of the next integer 値, the association of the frequency representative 値 can be established. The details of one of the preferred embodiments of this calculation are described in the specification disclosed here. This correction procedure must be performed before the compression procedure and before the amplitudes of all filters can be obtained. The preferred embodiment does not slow down the sampling rate, but stores all filter amplitudes 値 of the speech samples for about 10 to 15 seconds before the correction and compression processing, and the sampling rate is 10 to 15 samples per second. If the consideration of computer memory space is more important than the scanning speed, you can correct and compress between each scan to avoid a large amount of data storage memory requirements. Since small computers of ordinary price on the market are currently equipped with sufficient memory, the preferred and specific embodiments discussed herein can store all data and process the data later. 4Hickman200021tw; AND1P115.TW 47 This paper size applies to the Chinese National Standard (CNS) A4 specification (210X297 mm) (Please read the precautions on the back before this page) Threading · 548631 A7 ___B7____ V. Invention Description (Do?) Us Most of the verbal animal signals discussed include humans that contain the largest amplitude peaks that are unlikely to be at either end of the frequency domain. This peak can be judged by any generally simple numerical ordering algorithm as used in the present invention. Amplitude and frequency representatives are then placed in the third group of memory position groups that store the six peaks' amplitude and frequency. The highest frequency peaks above 8kHz are placed in the sixth memory location and marked as high frequency peaks. The lowest peak is placed in the first group of memories. Then choose another three groups of peaks from these two. After this compression processing, it becomes six peaks, each of which is represented by the amplitude and frequency, plus a period of time samples, such as ten times per second, a total of ten seconds of samples, and the unfiltered total signal energy amplitude to represent the voice signal . A total of 13,000 can be provided. Printed by the Consumer Cooperative of Intellectual Property Bureau of the Ministry of Economic Affairs. If the operator encounters unexpected noise interference, use the stop modification switch to modify the sample length switch to stop the operation. The calculation system can allow the sample length to change. The calculus system uses an average chirp that is not too sensitive to changes in the number of samples four or five seconds after the sound signal to allow this change. If possible, use a larger utterance sample, because the larger utterance sample captures the “style” of the speaker, and this style is usually evident in words within 10 to 15 seconds. The output of this compression function is sent to the storage algorithm system that combines the elements and forms the following: (a) The four voice qualities to be discussed next 4Hickman200021tw; AND1P115.TW 48 This paper is again applicable to the Chinese National Standard (CNS ) A4 specifications (210X297 mm) 548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of invention Kl) 値; (b) "Pause" or voiceless ratio of sound: (c) "Variability"- -The difference between the amplitude of each peak chirp between the current scan and the last scan; (d) Obtaining the ratio of the number of peak chirp changes greater than 0.4 between scans to the total number of sound scans to obtain "approximate syllable change"; And (e) "High Frequency Analysis"-the ratio of the number of audible scans with a non-zero chirp in the sixth chirp amplitude. There will be a total of 20 elements per scan. These contents are then transmitted to the range combination calculation system. The four voice qualities as elements are (1) the "spread range"-the average sample of the frequency between all scans above the maximum peak 値 amplitude representing the average 値 and the average 之下 below, (2) " "Balance"-the average sample obtained by dividing the average amplitude of peaks 4, 5, and 6 by the average of peaks 1 and 2 "(3)" Envelope smoothness "-the average of all scans with amplitudes above the maximum peak Taking the average sample of the maximum peak value, (4) "low envelope smoothness"-the average of all scans with amplitudes lower than the maximum peak value, divided by the maximum peak value. The range of sound styles is divided into "resonance" and "sound quality", which are combined by a calculation system using a matrix of coefficients acting on selected elements. The range of "speaking style" is divided into "change-monotone", "undulation-steady", "intermittent-continuous", "excited-soft" and "emotion-inhibited". The names of these five ranges respectively represent the bipolar style of each range. 'It is a calculation using the calculation system to act on a coefficient matrix calculation system of 15 elements out of 20 sound elements. 4Hickman200021tw; AND1PH5.TW 49 This paper scale applies to China Standard (CNS) A4 specification (X 297 public director) 548631 A7 B7___ 5. Description of invention (5D), please refer to Table 6 and specifications for details. The range of perceptual style is divided into "ecological-structure", "constant sensitivity", "other-self", "perceive-internal", "hate-love", "independent-dependent" and "perceptual-actual". The names of these seven perceptual ranges represent the poles of the range, respectively. They are calculated and combined by the calculation system to act on the coefficient matrix of the selected sound elements in speech and speech. For details, please refer to Table 7 and the specifications section. Commercially available standard computer keyboards or keypads allow users of the present invention to change all coefficients and redefine any combination of utterance, speech, or perceptual ranges for investigation. The selection switch can select to display any or all elements or ranges of the specified object sound sample. The digital processor is responsible for controlling the analog-to-digital conversion of the sound signal, and it also controls the recombination of speech elements into a range of speech, utterance, and perception digital 値. Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs. The microphone also interacts with a calculation system that combines voice, utterance, and perception range. Its importance is no less important than the operator's keypad input and the selection of the output screen and coefficient matrix selection. The output selector switch simply directs the output to any or all output sockets suitable for transmitting signals to commercially available standard displays, modems, printers, or a preset array of built-in light-emitting readout devices. Researchers can use this invention to develop group profile standards that list research results in publications based on occupation, dysfunction, homework, hobbies, culture, language, gender, age, animal species, and more. Alternatively, use 4Hickman200021tw; AND1P115.TW 50 This paper size applies Chinese National Standard (CNS) A4 specification (21 × X 297 mm) 548631 Printed by A7 B7, Consumer Cooperative of Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of invention (g \) You can compare your own puppet with other puppets issued by others or built into the machine. Now referring to the figure, use the microphone 1210 to input the speech into the speech analyzer, amplify the signal through the microphone amplifier 1211, or input the recording from the tape input socket 1212, using the pre-recorded speech pronunciation input. The input level controller 1213 can adjust the level of the voice signal sent to the filter drive amplifier 1214. The filter driver amplifier 1214 amplifies the signal and applies the signal to a V.U. meter 1215 to measure the correct operating signal level. The scan rate per second and the number of scans per sample are controlled by the operator using the scan rate and sample time switch 1216. The operator uses the sample start switch to stop and stop revision 1217 to start sampling. The revision function allows the operator to manually modify the set sampling time and stop sampling to prevent accidental sound interference, including simultaneous speakers, contaminating the sample. This switch is also responsible for connecting and disconnecting the microprocessor from the 110 volt power input pin. The output of the filter driver amplifier 1214 is also applied to a commercially available microprocessor-controlled filter bank and digitizer 1218. This device divides the audio signal of the sampling organization's audio range into 1/3 octaves, and divides each The voltage output of the filter is digitized. In a specific working embodiment of the present invention, 25 1/3 octave filters of the Eventide spectrum analyzer are used, and the center frequency of the filter ranges from 63 HZ to 16,000 HZ. In addition 4Hickman200021tw; AND1P115.TW 51 This paper size is applicable to Chinese National Standard (CNS) A4 specification (210X297 mm) 548631 A7 B7 5. Invention description (sq also uses AKAI microphone and recorder with built-in amplifier as filter banks And the input of digitizer 1218. The scan rate per second used by the filter bank is about ten scans per second. Other microprocessor-controlled filter banks and digitizers may work at different speeds. There are many micro-processing on the towel surface All the filters are suitable for controlling the aforementioned filter banks and digitizers. Regardless of the composite sound, the amplitude in the audio range in the 1/10 second time segment cannot be stable or flat, and there must be peaks and valleys. 1219 is the frequency representative of the peak of this signal, write down the amplitude 値 on both sides of the peak, and adjust its peak 往 to the adjacent filter with larger amplitude, which can make it more accurate. Because of the adjacent 1/3 octave filter Characteristic, the energy of a specific frequency will partially enter the adjacent filter according to the cutoff characteristics of each filter. In order to reduce this effect, there are only two adjacent filters The peak chirp filter's frequency will be regarded as the center frequency only when the amplitude is within 10% of its average chirp range. In order to ensure equal separation intervals, the linearization and normalization of chirps representing unequal frequency intervals, 25 filters The filters each specify a number of 1 to 25, which is used in the remaining steps of the processing program. Therefore, the 3,500 HZ difference between filters 24 and 25 becomes 1, which is also equal to that of the first and second filters. 17 HZ difference between each. In order to prevent more than five small blocks from appearing in each filter number, and to maintain an equal pitch between each small block of filter numbers 1 to 25, 4Hickman200021tw; AND1P115.TW 52 ( Please read the notes on the back before filling out this page}

, 11 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs This paper size is applicable to the Chinese National Standard (CNS) A4 specification (21〇 > < 297 mm)

548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the Invention (53) These small blocks are divided into a number of 0.2 pitches and are further specified as follows. If the amplitude difference between two adjacent filters and the peak chirp filter is greater than 30% of their average chirp, the number of the peak chirp filter is closer to the middle point of the next filter number than the peak chirp filter. If the adjacent filters represent higher or lower frequencies, respectively, this will increase the filter number of the peak-to-peak filter, such as filter number 6.0, to 6.4 or decrease to 5.6. If the larger adjacent filter amplitudes represent higher or lower frequencies, respectively, all other peaks will automatically have the filter number 値 +2 or -2 of the filter. After the aforementioned frequency correction 1220, the segmented and digitized spoken pronunciation 1219 is compressed, and only six amplitude peaks are stored, and the remaining peaks are discarded. The inventors found that as long as the following characteristics are noted, the six peaks 値 are sufficient to capture the style characteristics. At least one peak is close to the fundamental frequency; there can be only one peak in the area between the fundamental frequency and the peak frequency; the peak closest to the largest peak is retained; and the first two peaks stored above the largest peak are added At the 25th filter closest to the 16,000 HZ end or above 8 kHz, a total of six peaks are stored in the microprocessor memory. This ensures that the maximum peak is always the third peak stored in memory, the sixth peak stored at the same time can be used for high-frequency analysis, and the first peak is the lowest and closest to the fundamental frequency. After the signal is compressed to include the full-band amplitude 値, the filter number of the six peaks, and the thirteen 値 each of ten samples, a total of 10 seconds of samples (1300 値), the 1221 of Figure 12 begins to combine sound elements . 4Hickman200021tw; AND1P115.TW 53 This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) 548631 A7 B7 V. Description of the invention (5 {) The present invention uses the low frequency set and high frequency set in spoken pronunciation. Relationships provide elements of spoken pronunciation quality. On the other hand, comparisons of measurement combinations related to sound energy events such as pauses and decay rates determine elements of speech style. These speech style quality elements are obtained from the spectrum analysis of 1330, 1331, and 1332 in FIG. 13. Speaking style elements can also be derived from 1233, 1234, 1235, and 1236 in Figure 12 and the other four analysis functions shown in Table \ ^. The names of the stored sound quality analysis elements and their derivation are as follows: (1) Spectrum "distribution"-the average number of peak chirp filters above the maximum peak and the peak chirp filters below the maximum peak in each scan Sample average 値 of the number of filters between averages, Figure 13, 1330; (2) Spectral energy "balance"-a sample of the ratio of the sum of the sum of the amplitudes of the peaks above the maximum peak 与 to the sum of the amplitudes below the maximum 値Average, 1331; (3) Spectral envelope "flatness"-the ratio of the peak average amplitude (high) to the maximum peak value above the maximum peak value in each scan, and the peak amplitude (low) below the maximum peak value The ratio to the maximum peak 値, the calculated average of these two ratios in each sample, 1332. The stored speech style elements are named and derived as follows: (1) Spectrum variability The number difference between each peak 値 filter number in one scan and each corresponding peak 値 filter number in the next scan The six averages of a spoken sample and the six amplitudes of the six peaks, and the full-spectrum amplitude difference of each scan, yielding a total of 13 average samples, 1333; Full energy amplitude pause 4Hickman200021tw; AND 1P115 .TW 54 This paper size is applicable to China National Standard (CNS) A4 size (X297 mm) (Please read the precautions on the back before this page)

1T printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 _____B7 V. Description of the invention () (below the amplitude 値 two units) The number of scans and sound energy in the sample (3) the ratio of the number of syllable changes 値-the third peak change number 値 the ratio of the number of scans greater than 0.4 to the number of scans with sound in the sample, 1335; (4) and high Frequency analysis-The ratio of the number of sample scans to the total scans of the sixth peak with amplitude chirp. The method and device in the present invention divides the sound style into seven ranges, as shown in Table 6. These are deliberate decisions and are most sensitive to the range of seven perceptual or cognitive styles listed in Table 7. The program that associates sound style elements with speech, speech, and perceptual ranges for output (Figure 12, 1228) uses various equations to determine each range as a function of the selected sound style element, as shown in Figures 1330 to 1336. . Table 6 establishes the association between the speech style elements 1333 to 1336 and the speech style range in FIG. 13. Table 7 illustrates the relationship between the seven perceptual style ranges and the sound style elements 1330 to 1336. Similarly, the purpose of having an array of selective input coefficients containing zeros is to allow the device operator to switch or key in changes to these coefficients during the 1222 and 1223 investigations. A fast-acting operator can develop different ranges of perception, even personality or cognition, or coefficients (if he prefers the term), and development requires completely different coefficients. When developing, enter the required coefficient set, and note the 4Hickman200021tw; AND1P115.TW 55 associated with these coefficients. This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) 548631 A7 _ B7 —One ― __ — ~ 1 '—--------------- 一 " " " —V. Scope of Invention (fb) (1226). For example, some researchers may not want the other-self scope in Table 7 but instead replace it with an inward-outgoing user perception scope. After changing the other-self set coefficient, after experiments, find an acceptable high correlation between the selected weighted sound style element combination and its externally determined inward-outward range, the researcher can use the new inward-outward range Location, really rename it. This operation can make the sound element of this invention very sensitive to the user's inward-outward range, and the researcher's coefficient set can reflect the appropriate relationship. This can be done as long as many user-determined ranges are provided to bring them closer to practicality, and the present invention provides great productivity in the context of exploring, developing, or verifying new perceptual ranges related to sound style elements. Table 6 (Please read the notes on the back and the items on this page first) Binding and ordering (DSj) (l) coefficient elements (differences) in the range of speaking styles Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs Printed ESi (2) CSil CSi2 CSi3 CSi4 CSi5 No. 1 0 0 0 0 0 Amplitude 1 0 0 0 0 0 No. 2 1 0 0 0 1 Amplitude 2 1 0 0 1 0 4Hickman200021tw; AND1P115.TW 56 This paper size applies to the Chinese National Standard (CNS) A4 specifications ( 210X297 mm) 548631 A7 B7 V. Description of the invention uq) No. 3 amplitude 3 No. 4 amplitude 4 No. 5 amplitude 5 No. 6 amplitude 6 No. 7 amplitude peak 7 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 -1 0 1 (Please read the precautions on the back before this page) Printed by the Ministry of Intellectual Property Bureau's Consumer Cooperatives ## STR1 ## DS1 = Monotonic change DS2 = Changeable smooth DS3 = Intermittent coherent DS4 = Positive moderate DS5 = Emotional restraint. (2) Nos. 1 to 6 = peak-to-peak filter difference 1-6, amplitude 1 to 6 peak-to-peak amplitude difference 1-6. Amplitude 7 = full bandpass amplitude difference.

4Hickman200021tw: AND1P115.TW 57

Line 1T · This paper size applies Chinese National Standard (CNS) A4 specification (210 × 297 mm) 548631 A7 B7 V. Description of the invention (M) Table 7 (DPj) (l) coefficient element difference of perceptual style range EPi CPil CPi2 CPi3 CPi4 CPi5 CPi6 CPi7 Distribution 0 0 0 0 0 0 0 Balance 1 1 0 0 0 0 0 Env-H 0 1 0 0 0 0 0 Env-L 1 0 0 0 0 0 0 Number 1 0 0 0 0 0 0 0 Amplitude 1 0 0 0 0 0 0 0 No. 2 0 0 1 0 0 0 1 Amplitude 2 0 0 1 0 0 1 0 No. 3 0 0 0 0 0 0 0 Amplitude 3 0 0 0 0 0 0 0 No. 4 0 0 0 0 0 0 0 0 Amplitude 4 0 0 0 0 0 0 0 No. 5 0 0 0 0 0 0 0 1 Amplitude 5 0 0 0 0 -1 0 0 No. 6 0 0 0 0 0 0 0 Printed by the Employees' Cooperative of Intellectual Property Bureau of the Ministry of Economic Affairs 4Hickman200021tw; AND1P115.TW 58 This paper size applies Chinese National Standard (CNS) A4 specification (210X 297 mm) 548631 A7 B7 V. Description of the invention (q) Amplitude 6 0 0 0 0 0 0 0 0 Amplitude 7 0 0 0 1 1 0 -1 Pause 0 0 0 1 1 0 0 Peak 6 6 0 0 0 -1 -1 1 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs ## STR2 ## DPI = Ecological structure high-low; DP2 = unchanged High-low sensitivity; DP3 = others-self; DP4 = feeling-mind DP5 = Hate - Love; DP6 dependence - independence; DP7 = emotion - body. (2) Nos. 1 to 6 = peak 値 filters 1-6; amplitudes 1 to 6 = peak 値 amplitude difference 1-6; amplitude 7 full bandpass amplitude difference. The main result of the present invention that can be used by the user is the range 値 1226. The switch 1227 can be used to select and display on a standard light display, or it can be selected for use by a monitor, printer, modem, or other standard output device 1228. These can be used to determine how close the subject's voice is to any or all of the sound or perception range in the built-in, published, or personally developed controls or standards, and then help improve emotional recognition based on the results of the judgment. 4Hickman200021tw; AND1P115.TW 59 This paper size applies to Chinese National Standard (CNS) Α4 size (210 × 297 mm) (Please read the precautions on the back before IPN: page)

1T-line 548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the invention (b) In another exemplary embodiment of the present invention, the biological signal received from the user is used to assist in determining the user Emotions in speech. Improve the recognition rate of speech recognition systems by compensating for changes in user speech caused by factors such as emotions, anxiety, or fatigue. The utterance signal obtained from the user's pronunciation is modified by the preprocessor and provided to the utterance recognition system to improve the recognition rate. Discourse signals are modified based on biological signals that indicate the user's emotional state. Fig. 14 illustrates the structure of the speech recognition system in detail. The speech signal obtained from the microphone 1418 and the biological signal obtained from the biological monitor 1430 are received by the preprocessor 1432. The signal sent from the biological monitor 1430 to the preprocessor 1432 is a biological signal indicating the impedance between two points on the user's skin surface. The biological monitor 1430 measures the impedance using a contact 1436 connected to one finger of the user and a contact 1438 connected to the other finger. Biological monitors such as 63-664 biological feedback monitors sold by Radio Shack owned by Tandy Corporation, trade names (MICRONATA.RTM. BIOFEEDBACK MONITOR) can be used. These contacts can also be connected to other locations on the user's skin. When the user becomes excited or anxious, the impedance between two points 1436 and 1438 will decrease, and the monitor 1430 will detect this situation and generate a biological signal indicating that the impedance is reduced. The preprocessor 1432 uses the biological signal sent from the biological monitor 1430 to modify the utterance signal received by the microphone 1418. The utterance signal is modified to compensate for changes in the user's speech caused by changes caused by fatigue or changes in emotional state. For example, when the biological monitor 1430 shows that the user is in an excited state, the front processor 1432 can lower the tone of the utterance signal sent by the microphone 1418. Page) • Loading ·

1. The paper size of the 1T line is applicable to the Chinese National Standard (CNS) A4 specification (210X297 mm) 548631 A7 B7 Printed by the Intellectual Property Bureau of the Ministry of Economic Affairs g Consumption cooperatives 5. Description of the invention (1?) In a less excited state, such as when tired, the pre-processor 1432 can raise the tone of the utterance signal sent by the microphone 1418. The pre-processor 1432 then provides the modified utterance signal to the sound card 1416 in a conventional manner. The preprocessor 1432 communicates with the PC1410 using an interface such as an RS232 interface to perform operations such as initialization and calibration. The user 1434 communicates with the front processor 1432 by using the observation display 1412 and using the keyboard 1414 or the keypad 1439 or the mouse to input instructions. It is also possible to control the gain and / or frequency response of the microphone 1418 and pre-process the utterance signal using a biological signal. You can increase or decrease the microphone's gain or amplification in response to a biological signal. It is also possible to use a biosignal to change the frequency response of the microphone. For example, if the microphone 1418 is an ATM71 microphone of AUDIO-TECHNICA U.S., Inc., it can use a biological signal to switch the relatively flat response or the attenuation response. The low-frequency speech signal gain of the attenuation response is smaller. 0 If the biological monitor 1430 is a radio-shack upward reference monitor, the biological signal is a series of ramp-shaped signals, each ramp ^ period is about 0.2 msec. Figure 15 shows the biological signal. A series of slope-shaped signals 1542 are divided by time T. The amount of T time between ramps 1542 is related to the impedance between contacts 1438 and 1436. When the user is in a more excited state, the impedance between the contacts 1438 and 1436 will decrease, and the time T will also decrease. When the user is in a less excited state, the impedance between the contacts 1438 and 1436 becomes higher and the time T increases. 4Hickman200021tw; AND 1P115 .TW This paper size is applicable to Chinese National Standard (CNS) A4 specification (2 丨 0 ^ 7 public director) 61 548631 A7 B7 V. Description of the invention αι) The biological signal waveform input by the biological monitor may not be a series Slope signal. For example, the bio-signal may be an analog signal that changes its amplitude and / or frequency according to the measurement result of the bio-monitor, or it may be a digital signal based on the status measured by the bio-monitor. The biological monitor 1430 includes the circuit of FIG. 16 and generates a biological signal representing the impedance between the contacts 1438 and 1436. The circuit consists of two parts. The first part is used to sense the impedance between the contacts 1438 and 1436. The second part is an oscillator. The output connector 1648 will generate a series of ramp signals. The oscillation frequency is controlled by the first part. The first part controls the collector currents Ie, Q1 and voltages Ve, Q1 of the transistor based on the impedance between the contacts 1438 and 1436. In this embodiment, the impedance sensor 1650 is only the contacts 1438 and 1436 connected to the speaker's skin. Since the impedance change between contacts 1438 and 1436 is relatively slow compared with the oscillation frequency of the second part, compared with the second part, the 'collector current Ie, Q1 and voltage Ve, Q1 are almost stable status. Capacitor C3 makes these currents and voltages more stable. Part 2 is an oscillator. Reactance elements L1 and C1 control on and off of transistor Q3 to generate oscillation. When the power is just turned on ', the base current Ib, Q2 will be conducted, and Q2 will be turned on. Similarly, 1. , ^ Will also provide the base current Ib, Q3, so that the transistor Q3 is turned on. Inductor L1 has no current at first. When Q3 is turned on, the Vcc voltage is reduced by a small amount of saturated transistor 4Hickman200021tw; AND1P115.TW 62 This paper size applies to China National Standard (CNS) A4 specification (210X297 mm) (Please read the precautions on the back before this page)

—Equipment · .F Order Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 Δ7 Α7 Β7 V. Description of the Invention (C3) Pressing ¥ ^ 3 will be applied to both ends of L1. Therefore, the current Iu increases with it. As the current L increases, the current Iel passing through the capacitor C1 also increases. Because the currents ^, (^ are completely fixed, increasing the current I will reduce the base current IB, Q2 of the transistor Q2. This will cause the currents Ie, Q2, \ (^, and kQ3 to decrease. So there will be more The current L flows through the capacitor Cl ′ to reduce the current 1., (3). This feedback effect causes the transistor Q3 to be turned off. At the end, the capacitor C1 is fully charged, and the currents L and Iel are reduced to zero, so that the current Ie Q1 can draw currents Ib, Q2 again, so that transistors Q2 and Q3 are turned on, and the oscillation cycle is restarted. Current 1., ^ will change with the impedance between contacts 1438 and 1436. This current controls the output signal. Periodic frequency. As the impedance between contacts 1438 and 1436 increases, the time T between ramp signals decreases. When the impedance between contacts 1438 and 1436 increases, the time T between ramp signals increases. Intellectual Property of the Ministry of Economic Affairs The power of the printed circuit of the Bureau ’s consumer cooperative is supplied by a three-volt battery 1662, which is connected to the circuit via a switch 1664. There is also a variable resistor 1666 in the circuit to set the operating point of the circuit. It is best to change the variable Resistance 1666 is set at Near the middle point of its working range. Then the circuit will change according to the previous description based on the impedance between the contacts 1438 and 1436. 4Hickman200021tw; AND1P115.TW 63 This paper standard applies Chinese National Standard (CNS) Α4 Specifications (210X297 mm) 548631 Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of invention αψ) The circuit also contains a switch 1668 and a speaker 1670. When no matching connector is inserted in the connector 1648, the switch 1668 will The output is sent to the speaker 1670, but not to the connector 1648. Figure 17 is a block diagram of the preprocessor 1432. The analog-to-digital (A / D) converter 1780 receives the speech or pronunciation signal input from the microphone 1418, and the analog-to-digital (A / D) converter 1782 receives the biological signal input from the biological monitor 1430. The signal output from the A / D 1782 is sent to the microprocessor 1784. The microprocessor 1784 monitors the signal input from the A / D 1782 to determine the digital signal processor ( DSP) device 1786 should take action. Microprocessor 1784 uses memory 1788 to store programs and perform cut and paste operations. Microprocessor 1784 uses RS232 interface Communicates with the personal computer 1410. The software controlling the interface between the PC 1410 and the microprocessor 1784 may be executed on the personal computer 1410 with a multi-application environment, using, for example, a brand name (WINDOWS) program sold by Microsoft Corporation. Software package. The output of DSP 1786 is converted back to analog signal by digital-to-analog converter 1790. After the DSP 1786 modifies the signal of the A / D 1780 under the instruction control of the microprocessor 1784, the output of the D / A converter 1790 is transmitted to the sound card 1416. The microprocessor 1784 can use a common microprocessor on the market, such as the microprocessor of Intel Corporation, and the DSP1786 can also use a common digital signal processing chip on the market, such as the TMS320CXX device system of Texas Instruments [J. The biological monitor 1430 and the front processor 1432 can be placed on the same interface. 4Hickman200021tw; AND1P115.TW 64 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ---------- --Installation ------- Order ------ (Please read the precautions on the back before filling out this page) Printed by the Intellectual Property Bureau Employee Consumer Cooperative of the Ministry of Economy 548631 A7 B7 V. Invention Description (0 ) Card, insert the empty slot in the personal computer 1410. The personal computer 1410 can perform the functions of the microprocessor 1786 and the digital signal processor 1786 without the need for dedicated hardware. The microprocessor 1784 monitors the biological signal of the A / D 1782 to determine what action the DSP 1786 should take. When the signal output by the A / D 1782 indicates that the user is in a more excited state, the microprocessor 1784 will notify the DSP 1786 to process the A / D 1780 by lowering the tone of its utterance signal. When the output signal of A / D 1782 shows that the user is in a less active or tired state, the microprocessor 1784 instructs the DSP 1786 to raise the tone in the utterance signal. The DSP 17S6 will build a speech model to modify the tone of the speech signal. The DSP then uses the model to re-establish the utterance signal with the modified tones. The discourse model is built using one of the very well-known linear predictive coding techniques. Analog Devices, Inc.'s application book "Digital Signal Processing Applications Using the ADSP 2100 Family" has pages 355-372 describing this technique. Who is the publisher of this book? ^! ^ 06-Hall, Englewood Cliffs, N.J., 1992. This technique involves making a speech signal into a FIR (Finite Impulse Response) filter model with a time-varying coefficient, using a series of pulses to excite the filter. The time T between pulses is some tone or fundamental frequency. The coefficient of time variation can be calculated using techniques such as Levinson-Durbin cycles, which are described in the aforementioned book published by Analog Device, Inc. The inter-pulse time T of the series of pulses constituting the excitation filter can use 4Hickman200021tw of John D. Markel; AND1P115.TW 65 This paper size is applicable to China National Standard (CNS) A4 (210 χ 297 mm) (Please read the back first Please pay attention to this page and fill in this page) -------- Order · ---- 111 · 548631 A7 B7 Printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs V. Invention Description SIFT (Simple Inverse Filter Tracking) Calculations such as calculations, this method is described in the chapter "M The SIFT Algorithm for Fundamental Frequency Estimation" in John D. Markel's IEEE Transactions on Audio and Electroacoustics, Vol. AU-20, No. 5, December, 1972. DSP I786 The time T between pulses is changed when the FIR filter is excited to reconstruct the speech signal, thereby modifying the fundamental frequency tone of the speech signal. For example, if the time T between pulses is reduced by 1%, the tone is increased by 1%. Note that, The speech signal can be modified without changing the tone. For example, the tone, amplitude, frequency, and / or signal spectrum can be modified. Part or all of the signal spectrum can be attenuated or amplified In addition to monitoring biological signals showing impedance between two points on the user's skin, other biological signals can also be monitored. Signals showing automatic activity can be used as biological signals. Signals showing automatic activity can be used, such as blood pressure, pulse number, brain Wave or other charged activity, pupil size, skin temperature, permeability or reflectivity of specific electromagnetic wavelengths or other signal representations showing the user's emotional state. \ Λ 18 is a microprocessor 1784 used to instruct the DSP 1786 to communicate with biological signals Relevant time T changes the tone modification curve of the tone of the speech signal. The horizontal axis 1802 represents the time between the biological signal slopes 1442, and the vertical axis 1804 represents the percentage change of the tone caused by the output of the DSP 1786. 4Hickman200021tw; AND1P115.TW 66 This paper scale applies China National Standard (CNS) A4 Specification (210 X 297 mm) ------------------- Order · -------- (Please read the Note: Please fill out this page again) 548631 a? B7 V. Description of the invention uq) Figure 19 is a flowchart of instructions executed by the microprocessor 1784 to establish the working curve shown in Figure 18. After starting, execute Step 1930 is performed to establish a straight line that is the same linear as the 1802 axis. This line represents that no pitch change has been added to all T 値 of the biological signal. After step 1930, a decision step 1932 is executed, and the microprocessor 1784 determines whether a modification instruction input from the keyboard 1414 or the keypad 1439 is received. If no modification instruction is received, the microprocessor 1784 executes a loop and waits for the modification instruction. If a modification instruction is received, step 1934 will be performed to determine TT = Ί ^ η 値 to be used to establish a new reference point Ref1. Trefl 値 is equal to T current 取得 obtained from biological signals. For example, Trefl may be equal to 0.6 msec. After Trefl 値 is determined, the microprocessor 1784 executes step 1938 and asks the user to pronounce it in order to obtain a tone sample in step 1940. Since tone samples can be used as the basis for the percent change in tone indicated along the 1804 axis, it is best to obtain a tone sample. In step 1942, the microprocessor 1784 instructs the DSP 1786 to increase the tone of the speech signal by an amount equal to the current tone change associated with the Refl point plus a five percent increment; however, a smaller or larger amount may be used. In increments. (At this time, the tone change associated with the Refl point is zero. Please refer to step 1930.) In step 1944, the microprocessor 1784 will ask the user to say a few instructions to the speech recognition system to perform a recognition test to determine whether it is acceptable. Recognition rate. After the user finishes the test, he can use the keyboard 1414 or the keypad 1439 to input an instruction such as "end" to tell the microprocessor 1784 that the test has been completed. 4Hickman200021tw; AND1P115.TW 67 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) (Please read the note on the back? Matters before filling this page) ·! ------ t Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the invention (G?) After the microprocessor 1784 performs step 1944, Then execute step 1946 to instruct the DSP 1786 to reduce the tone of the input speech signal by a 5% reduction with the tone change associated with the Refl point; a smaller or larger decrease can also be used (please note that during step 1930 After that, the tone change associated with the Refl point is zero). In step 1948, the microprocessor 1784 will ask the user to perform another speech recognition test and enter the “nend” command after the test is completed. In step 1950, the microprocessor U84 will ask the user to choose the first or second test. Indicate which test has better recognition ability. Step 1952 uses the result selected by the user to select step 1954 or 1956. If test 1 is selected as the best test, step 1956 will be performed, and the new percentage change set with the Refl point will be set Is equal to the previous Refl point 値 plus five percent or the increment used in step 1942. If test 2 is selected as the best test, step 1954 is performed, and the change in the percentage of connection of Refl is set equal to the old Refl 値Subtract 5 percent or the decrement used in step 1946. The percentage change in the decision and knot will create a new reference point Refl. For example, if test 1 is selected as the best test, the Refl point will be located at 1858 in Figure 18. Create After the position of the newly established Refl point 1858, a straight line 1860 is established in step 1962. The straight line 1860 is used to calculate different T 値 percentage changes in the biological signal Initial Tone Modification Line. The slope of this line can be specified to increase by 5% per millisecond at the beginning, but other slopes can also be used. After establishing this initial modification line, the microprocessor 1784 will enter the waiting loop for execution Steps 1964 and 1966. In step 1964, the microprocessor 1784 4Hickman200021tw; AND1P115.TW 68 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ----------- Packing -------- Order --- (Please read the notes on the back before filling out this page) Gift _ Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Invention Description aq) Will check and modify Instruction, and the deactivation instruction is checked in step 1966. If no modification instruction is received in step 1964, the processor will go to step 1966 to check the deactivation instruction. If no deactivation instruction is received, the microprocessor will return to step 1964, If a disable instruction is received, the microprocessor will execute step 1930 to set the tone changes of all T 等于 in the biological signal to be equal to zero. The processor will stay here to check the circuit of the modification and disable instructions until the The user is not satisfied with the recognition rate generated by the pre-processing of the utterance signal made by the curve 1860. If a modification instruction is received in step 1964, step 1968 is performed. Step 1968 determines the 値 of T and checks whether 检查 of T is equal to or It is almost equal to eight of the Refl point. If T of T is the same as Refl, step I942 is performed. If T of T is not the same as Refl, step 1970 is performed. Step 1970 creates a new reference point Ref2 of 1 \ ^ 値. For the sake of explanation, we assume Tref2 = 1.1 m sec. Referring to Figure 18, this will establish point 1872 on line 1860 as Ref2. In step 1974, the microprocessor 1784 instructs the DSP 1786 to increase the tone change associated with the Ref2 point by 2.5% (other percentages may also be used). Step 1976 asks the user to perform a recognition test, and enters an ' end " command upon completion. In step 1978, the microprocessor 1784 instructs the DSP 1786 to reduce the tone of the utterance signal, and the reduction is caused by a reduction of the tone change related to ReO by 2.5%. Step 1980 will then ask the user to perform the recognition test, and also enter the "end" command after completion. Step 1982 will ask the user to indicate that the results of the first or second test are better. In step 1984, the microprocessing 1784 will be based on 4Hickman200021tw; AND1P115.TW 69 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) ----------- Aw — (Please read the precautions on the back first (Fill in this page again) Order --- Suspect-548631 A7 B7 Printing of clothing by the Consumer Property Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the invention () The best test selection results are performed in different steps. If test 1 is selected, step 1986 will be performed. If test 2 is selected, step 1988 will be executed. In step 1986, the microprocessor 1784 will set the percentage change associated with the Ref2 point to the Ref2 related old value plus 2.5% or the increment used in step 1974. In step 1988 The percentage change of the Ref2 correlation is set equal to the old correlation of Ref2, minus 2.5% or the reduction used in step 1978. After completing step 1986 or I988, step 1990 will be executed. Step 1990 will create a new Adjust the modification line. The new modification line uses the correlation line of Refl and the new correlation point of Ref2. For example, suppose the user selects test 1 in step 1984. The new Ref2 correlation point will be point 1892 in Figure 18. Now the new tone conversion The line has become a straight line 1898 passing through two points of 1892 and 1858. After the microprocessor 1684 has performed step 1990, it will return to the loop operation composed of steps 1964 and 1966. Please note that the linear modification line is used here, but it is also possible Use a non-linear modification line. This is done by using points 1858 and 196 to establish the slope of the line to the right of connection point 1858, and using another reference point of the left to connect point 1858 to establish the slope of the line to the left of connection 1858. You can also add a maximum The positive and negative limits of the percentage tone change. When the tone modification line approaches these limits, it can be approached gradually, or it can change suddenly at the point in contact with the limit. Alternatively, you can use a fixed modification curve, such as curve 1800, and adjust the variable resistance 1666 Until an acceptable recognition rate is reached. 4Hickman200021tw; AND 1P115.TW 70 This paper size applies Chinese national standards (CNS) A4 specification (210 X 297 mm) ------------ install --- (Please read the precautions on the back before filling this page) Order -------- -Li_ 548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the Invention (rjl) Voice Message Transmission Figure 20 illustrates a specific embodiment of the present invention, which is managed according to the emotional characteristics of the voice message Voice message. Job 2000 receives some voice messages transmitted over a telecommunications network. Assignment 2002 stores voice messages on a storage medium, such as the recorder or hard drive mentioned earlier. Assignment 2004 will determine the emotion associated with the voice signal in the voice message. Emotions can be judged using any of the methods mentioned earlier. Then organize the voice message in homework 2006 based on the judged emotions. For example, messages that show negative emotions such as sadness, anger, or fear in the voice can be grouped in the mailbox and / or database. Assignment 2008 has access to organized voice messages. The voice message will follow the call. You can choose to organize voice messages with similar emotions together. You can also choose to organize voice messages immediately after receiving a call. It is best to identify a way to organize your voice messages to assist in accessing your organization's voice messages. In addition, it is also best to judge at least one feature from the voice message according to the previous discussion. In the exemplary embodiment of the exemplary voice message transmission system made according to the present invention, the tones and LPC parameters (usually also including other excitation information) are encoded for transmission and / or storage, and then decoded to provide and original speech input. Similar answer. 4Hickman200021tw: AND1P115.TW 71 This paper size is applicable to China National Standard (CNS) A4 specification (210 X 297 mm) ----------- · Installation -------- Order --- ------ (Please read the precautions on the back before filling out this page) 548631 A7 B7 Printed by the Consumers' Cooperative of Intellectual Property Bureau of the Ministry of Economic Affairs V. Description of the invention (f / >) This invention is particularly relevant for analysis Or linear predictive coding (LPC) systems (and methods) that encode human speech signals. In the LPC model, each sample in the series of samples is usually modeled (simplified model) as a linear combination of the previous samples plus an excitation function:

Sk = t ^ k-J + uk 7 = 1 where Uk is the LPC residual signal. That is, Uk represents the residual information in the input speech signal that was not predicted by the LPC model. Here only N previous signals are used for prediction. The order of the models (usually around 10) can be increased to provide better predictions, but in the application of general discourse models, some information must be stored in the residual signal uk. In the general framework of the LPC model, many specific speech analysis implementations can be selected. Many of them must judge the tone of the input speech signal. That is, in addition to the formant frequency that is actually consistent with the resonance of the vocal tract, human speech also includes a tone that is tuned by the speaker and corresponds to the frequency of the laryngeal adjustment of the airflow. In other words, human vocalization can be regarded as the excitation function applied to the auditory passive filter. This excitation function usually appears in the LPC residual function, and the characteristics of the passive auditory filter (resonance characteristics of the oral cavity, nasal cavity, and chest cavity, etc.) ) Is shaped by LPC parameters. When there is no sound, the tone of the excitation function is not clear, and its optimization model is broadband white noise or pink noise. 4Hickman200021tw; AND1P115.TW 72 ^ The paper size is applicable to the Chinese National Standard (CNS) A4 (210 X 297 mm). -------- Order --------- (Please read the back first Please pay attention to this page before filling in this page) 548631 A7 B7 V. Description of invention (rj3) It is estimated that the period of ten sounds and g weeks is not a trivial matter and many problems will be encountered. One of them is that the first formant usually appears near the tone frequency. Therefore, the tonal estimation is often performed using the LPC residual signal, because the LPC estimation process is actually expanding the channel resonance from the excitation information, so that the residual signal contains relatively few channel resonances (formants) and contains relatively more excitations. Information (tone). However, this residual tone estimation technique has its own difficulties. The LPC model itself usually introduces high-frequency noise into the residual signal. The spectral density of each part of this high-frequency noise is higher than the actual tone to be detected. One of the difficult solutions is to low-pass filter the residual signal around 1000 Hz. This removes high-frequency noise, but at the same time removes high-frequency energy present in the silent area of the pronunciation, making the residual signal almost useless for applications that determine pronunciation. The overriding principle of voice messaging applications is the quality of the voice played. Prior art systems had many shortcomings in this regard. In particular, many of them were unable to accurately detect tone and pronunciation problems in input voice signals. It is often possible to erroneously estimate the duration of the tone to be twice or half the actual chirp. For example, when using the correlation method, if the correlation of period P is good, the correlation of period 2P must also be good, and the signal will have a good correlation during period P / 2. However, this double or halving error can degrade the sound quality and is very disturbing. For example, halving errors during tones often results in harsh creaks' and doubling errors often results in rough sounds. In addition, the doubling or halving during the tone period is likely to occur intermittently, making the synthesized speech easy 4Hickman200021tw; AND1P115.TW 73 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ---- ------- Installation (please read the precautions on the back before filling this page) ---- Order --------- Intellectual Property of the Ministry of Economic Affairs Bureau Employee Consumption Co-operative Printing Clothing Intellectual Property of the Ministry of Economic Affairs Printed by the Bureau's Consumer Cooperatives 548631 a? B7 5. Description of the Invention (ah) Intermittent husky or friction occurs. The present invention uses an adaptive filter to filter the residual signal. Use the first reflection coefficient (the kO of the speech input has a single-pole time-varying filter to remove high-frequency noise during the vocal period of the speech, but retain the high-frequency information during the silent period. Then use the adjusted filter Residual signals are used as input for tonal decisions. To make better vocal / silent decisions, high frequency information must be preserved during periods of silent speech. That is, if a strong tone is not found, it is usually judged to be a "silent" pronunciation That is, when the residual signal has no correlation hysteresis, it will provide a high normalized correlation. However, if only the low-pass filtered part of the residual signal during the silent period is tested, some fragments of the residual signal will show a false correlation. The danger is that the clipped residual signal generated by the fixed low-pass filter of the previous technology does not contain data that can reliably show that there is no correlation during the silent period, and must rely on the extra bandwidth provided by the high-frequency energy during the silent period. Reliably exclude false correlation lags that may be found. Improved pitch and pronunciation decisions for speech The information transmission system is particularly important, and of course it is also beneficial to other application environments. For example, a document recognition system using tone information must have a good tone estimation process. Similarly, tone information is sometimes used to identify the speaker, especially On the telephone line where high-frequency information will be partially lost. In addition, we also hope that the identification system in the future is 4Hickman200021tw; AND IP 115.TW 74 This paper size applies to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) --- -------- Aw ^ -------- Order --------- (Please read the notes on the back before filling this page) 548631 A7 ______ B7 V. Description of the invention ( #) The system can take into account the syntactic information represented by the tone. Similarly, some advanced speech recognition systems, such as speech input text systems, also hope to have good pronunciation analysis capabilities. V The first reflection coefficient h is about and high / The low frequency energy ratio is related to the signal. Please read " Design of a Robust Maximum Likelihood Pitch Estimator for Speech and Additive Noise, f, Technical Note, 1979--28, Lincoln Labs, June by R.J. McAulay 11, 1979, mentioned here for reference only. When h is close to -1, the low-frequency energy in the signal is more than the high-frequency energy. When h is close to 丨, the situation is reversed. Therefore, using h to determine the unipolar solution At the pole, the residual signal during vocalization will be low-pass filtered, and the residual signal during silent vocalization will be high-pass filtered. That is to say, the formant frequency will be excluded when calculating the tone during vocalization, and it is necessary during the silent period. High-bandwidth information is retained to accurately detect situations where there is no tone correlation. It is best to use post-processing dynamic programming techniques to provide the best pitch and best pronunciation decisions, that is, to track the pitch and pronunciation of each frame one by one, accumulating a series of frame pitches for each track The cumulative loss of / voice decisions, find the trajectory that provides the best tone and pronunciation decisions. Cumulative loss is obtained by using the feature that frame errors will continue to the next frame. Frame errors not only cause a large amount of deviation during the tone period from frame to frame ', but also cause the situation that the correlation of the tones is "good" and relatively poor, and cause changes in pronunciation decisions in the frequency spectrum with relatively little change in the front and back frames. The last news 4Hickman200021tw; AND1P115.TW 75 This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm) (Please read the precautions on the back before filling this page). 11111111 Employees of Intellectual Property Bureau, Ministry of Economic Affairs Printed by the Consumer Cooperative 548631 A7 B7 V. Description of the Invention (u) The characteristics of the wrong frame transition will force the pronunciation transition towards the point of maximum spectrum change. The voice message transmission system of the present invention includes providing a speech input signal that is displayed in time sequence si for the LPC analysis block. A variety of conventional techniques can be used to perform the LPC analysis. The result is a set of LPC parameters and residual signal W. Many well-known references refer to the background of various methods for LPC analysis and extraction of LPC parameters, including Markel and Gray's Linear Prediction of Speech (1976) and Rabiner and Schafer's Digital Processing of Speech Signals (1978) and the references cited therein. Reference, mentioned here for reference only. In the presently preferred embodiment, the analog speech waveform is sampled at a frequency of 8 KHz with 16-bit accuracy to generate the input timing Si. Of course, the present invention is completely independent of the sampling rate or accuracy used, and can be applied to utterances sampled at any rate, and to any level of accuracy. In the presently preferred embodiment, the LPC parameter set used includes many reflection coefficients h, while a tenth-order LPC model is used (that is, only reflection coefficients k! To k1Q are extracted, and no more coefficients are extracted). However, as is familiar to those skilled in the art, other models or other equivalent LPC parameter sets can also be used. For example, LPC predictor coefficients ~ or impulse response estimation ek can be used. However, the most convenient is the reflection coefficient k / 4Hickman200021tw; AND1P115.TW 76 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) (Please read the precautions on the back before filling this page) Order- -Intellectual Property Bureau, Ministry of Economic Affairs, Employee Cooperatives, India, 548631 A7 B7 5. Invention description) In the current preferred embodiment, the reflection coefficient is extracted according to the Leroux-Gueguen program. For the description of this program, refer to kiEE. on Acoustics, Speech and Signal Processing, p. 257 (June 1977) and other documents, mentioned here for reference only. However, skilled technicians can also use other common algorithms, such as Durbin's algorithm, to calculate the coefficients. The calculation of LPC parameters is usually accompanied by a residual signal Uk. However, if the parameter is calculated using a method that does not automatically generate the by-product uk, as long as the finite impulse response digital filter is constructed using the LPC parameter, the filter will directly calculate the residual signal uk using the input timing sk. The residual signal timing uk then goes through a very simple digital filtering operation based on the LPC parameters of the current frame. That is, the input signal Sk will be at a certain sampling rate, such as 8 KHz, and each sample will change its timing once. However, LPC parameters are usually recalculated only once during each frame at a frame frequency of 100 Hz. The period of the residual signal nk is also equal to the sampling period. Therefore, the digital filter based on the LPC parameter is best not adjusted in each residual signal ixk. In the presently preferred embodiment, before the new LPC parameter 値 is generated, about 80 residual signal timings uk pass through the filter 14, so the new characteristics of the filter 14 will be implemented. More specifically, the first reflection coefficient h is 4Hickman200021tw; AND1P115.TW 77 from the LPC analysis section 12. This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) (please read the note on the back first) ? Please fill in this page again) Install ---- tr --------- Printed by the Consumers' Cooperative of Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 548631 V. Extraction of the LPC parameters provided by the description of the invention (Γ ^) . The LPC parameter itself is the reflection coefficient h, and only the first reflection coefficient h needs to be queried. However, when other LPC parameters are used, the parameter change that produces the first reflection coefficient is usually very simple, such as k ^ a ^ ao. Although the present invention preferably uses the first reflection coefficient to define a single-pole adaptive filter, the present invention The scope of application is not limited to this main preferred embodiment. That is, the filter does not have to be a single-pole filter, and can be constructed as a more complex filter with one or several poles and / or one or several zeros, some or all of which can be adjusted and changed according to the present invention. Also note that the characteristics of the adaptive filter need not be determined by the first reflection coefficient h. We know that there are many other equivalent LPC parameter sets. The parameters in other LPC parameter sets can also provide the required filtering characteristics. Especially in any LPC parameter set, the lowest-order parameter may usually provide information about the overall spectrum shape. Therefore, the adaptive filter determined according to the present invention can use 31 or define poles, which can be single-pole or multi-pole, or can be used alone or with other zero and / or poles. In addition, the poles (or zeros) defined by the LPC parameter adaptation do not have to be completely consistent with this parameter, as in this preferred embodiment, their size or phase can be changed. 4Hickman200021tw; AND1P115.TW 78 This paper size applies to Chinese National Standard (CNS) A4 (210 X 297 mm) -------- ^ -------- (Please read the precautions on the back before (Fill in this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Description of Invention (M) Therefore, the single-pole adaptive filter will filter the residual signal timing Uk to generate the filtered timing U'k. As discussed earlier, the high-frequency energy of the time sequence u′k after this filtering will be greatly reduced in the vocal segment, but almost the entire bandwidth can be reserved in the silent segment. The filtered residual signal U'k is then subjected to further processing to extract candidate tones and pronunciation decisions. There are many ways to extract tone information from the residual signal, and all of these methods can be used. Many of these methods are described in the previously mentioned books by Markel and Gray. In the current preferred embodiment, candidate tones are obtained by finding the peaks in the normalized correlation function of the filtered residual signal, which are defined as follows: ^ ujuj-k

Ck: 1/2 1/2 forkm < kr Σ Σ k -----------. Equipment -------- Order --------- (Please read first Note on the back, please fill in this page again.) The printed clothing u'j of the employee's consumer cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs is a filtered residual signal. The number of samples (80 in the preferred embodiment) defines the number of samples to be associated. Candidate tone 値 is defined by the time when hysteresis k * is the largest at C (k *). C (k) has no vector 値 and is used to define the "goodness" of each candidate k *.

4Hickman200021tw; AND1P115.TW 79 This paper size is applicable to Chinese National Standard (CNS) A4 (210 X 297 mm) 548631 Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. The description of the invention (π) can be chosen in good quality The critical 値 cmin is used in the measurement of c (k). The maximum C (k) at the local end that does not exceed the critical 値 Cmin will be ignored. If k * 値 does not make c (k *) greater than cmin, this frame must be silent. It is also possible not to use the goodness threshold 値 cmin, as long as the normalization autocorrelation function 1112 is controlled to select some candidate 値 with the best 良好, such as the 16 candidate tone periods 最大 with the largest c (k) 値. In a specific embodiment, there is no criticality at all, which is imposed on goodness C (k), and no pronunciation decision is made at this stage. Instead, choose 16 tone periods 値 kk ', k * 2, etc. and their corresponding goodness 値 (Qk,)). In the presently preferred embodiment, even if all C (k) 値 are extremely low, the pronunciation decision will not be made at this stage, but the pronunciation will be determined in the subsequent dynamic programming steps discussed below. In the presently preferred embodiment, many candidate tones are found based on the peak-to-peak search algorithm. That is, the relationship between the "goodness" quality C (k) and the candidate tone period is tracked. Each local maximum 値 will be regarded as a possible peak 値. However, after the function finds a fixed number, it will confirm that there is no peak at the local end of the recognition. This confirms the largest chirp at the local end and then provides one of the candidate chirps during the tones. After finding each crest candidate in this way, the calculation system then looks for troughs. That is to say, each local minimum is regarded as a possible trough, but it will not be confirmed as a trough until the function is raised by a predetermined constant. Bottom will not decide individually, 4Hickman200021tw; AND1P115.TW 80 This paper size applies to China National Standard (CNS) A4 specifications (210 X 297 mm) ------------ installation ---- ( Please read the notes on the back before filling this page) · 11111111 548631 A7 B7 V. Description of the invention (g () But after a confirmed peak, there must be a confirmed trough to find the next new peak. In the current preferred embodiment, the definition of goodness 値 is limited to +1 or -1, and the constant 确认 required to confirm the peak or trough is set to 0.2, but it can be greatly changed. Therefore, this stage provides 0 to 15 Tone candidate 値 is used as an output. In the current preferred embodiment, the candidate tone period set provided in the previous step is then provided to the dynamic programming algorithm system. This dynamic programming algorithm system tracks the tone and pronunciation decisions for Each adjacent frame in its adjacent context provides tonal and pronunciation decisions. With candidate tones 値 and its good 値 c (k), then dynamic programming is used to obtain the most inclusive of each frame. The best tone contour determined by pronunciation. Dynamic programming needs to analyze many speech frames in a discourse before determining the tone and pronunciation of the first frame in the segment. In each frame of the discourse segment, each tone will be The candidate tone is compared with the tone candidate reserved from the previous frame. Each retained tone candidate in the previous frame contains a cumulative loss, and every comparison between each new tone candidate and any reserved tone candidate. There will also be a new gap divisor. Therefore, each tone candidate in the new frame will have a representative minimum loss that corresponds to the tone candidate reserved in the previous frame. Calculate the minimum cumulative loss for each new candidate. At the same time, with the candidate 値, there is also a backward indicator of its cumulative loss and pointing to the best matching 値 in the previous frame. The backward indicator defines a curve to generate a curve at the planning rate as the last message 4Hickman200021tw; AND1P115.TW 81 copies Paper size applies to China National Standard (CNS) A4 (210 X 297 mm) (Please read the note on the back first) Please fill in this page for the matters) Packing -------- Order --------- Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Description of Invention (J >) Cumulative Loss in the Box The cumulative loss listed in 値. Select the curve with the smallest cumulative loss to get the best curve for any given frame. The silent state is defined as the tone candidate for each frame. The loss function preferably contains pronunciation information to make the dynamic The programming strategy can naturally generate pronunciation decisions. In the current preferred embodiment, the dynamic programming strategy has a width of 16 and a depth of 6. That is, each frame finds 15 (or fewer) candidates. The "silent" decision (designated as a zero-tone period for convenience) as a possible tone period, the first six frames will retain all 16 candidates (and their goodness). Printing of clothing by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs The tone and pronunciation will only be determined by the oldest frame included in the dynamic programming algorithm system. That is, the tones and pronunciation determine the candidate tones that will accept the FK-5 frame with the smallest current curve cost. That is, in the recent frame? 16 Of the 16 (or fewer) curves ending, the candidate tone with the lowest cumulative curve cost in frame FK is the best curve. Then follow and use this best curve to determine the tone / pronunciation of frame FK-5. Note that the next frame (Fk-4, etc.) will not make a final decision on the tone candidate, because after estimating more frames, the best curve is no longer the best. Of course, anyone who is familiar with digital optimization knows that at other times, such as the last second frame retained in the buffer area, the final decision of the dynamic programming calculation can be made. In addition, the width and depth of the buffer can also vary significantly. For example, you can evaluate up to 64 tone candidate chirps, or you can only evaluate two; only the previous frame can be kept in the buffer, or the first 16 can be kept. 4Hickman200021tw; AND1P115.TW 82 This paper standard applies to Chinese National Standards (CNS) A4 specifications (210 X 297 mm) 548631 A7 Printed by B7 of the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs. 5. Description of the invention (¾¾) More than one frame. Those skilled in the art can make other modifications and changes. The dynamic programming algorithm is defined by a transition error between the candidate period 値 in one frame and the candidate period 其他 in the other frame. In the presently preferred embodiment, the definition of this transition error is the sum of three parts: the error Ep caused by the pitch deviation, the error Es caused by the "good" low pitch candidate, and the error Et caused by the pronunciation transition. The tone deviation error Ep is a function of the current tone period and the previous tone period. If both frames have an audio signal, the function formula is as follows:

Ad + Bp, tau In tdUp Ep-mm Ad + βρ, tau In t (XUp + Bp \ n2 > Ad + Bp Λ tau In, taup + ln (l / 2) otherwise Ep = BP times DN; tau is the current frame In the candidate tone period, taup is the retention period of the previous frame whose transition error is to be calculated, and BP, Ad, and DN are all constants. Note that the minimum function includes the assumption that the tone period is doubled and the tone period is halved. The present invention It is not necessary to have this assumption in China, but we believe that this assumption will help. Of course, similar assumptions can also be included in the triple tone period, etc. 4Hickman200021tw; AND1P115.TW 83 This paper standard applies Chinese National Standard (CNS) A4 specification (210 X 297 mm) ----------- install -------- order --------- (Please read the precautions on the back before filling in this Page) Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 5. The invention states that the pronunciation error Es is a function of the“ goodness ”値 C (k) of the current frame tone candidate 考. The silent candidate 値 will definitely Contains 16 or less tone period candidates to be considered in each frame, its goodness C (k) is set equal to the maximum C (k) of candidate chirps during all 15 other tones in the same frame. When the candidate chirp is voiced, the pronunciation state error Es is ES = BS (RV-C (tan), otherwise Is Es = Bs (C (taU) -Ru), where C (taii) is the "goodness" corresponding to the current tone candidate 値,

Bs, and Ru are constants. Pronunciation transition error Eτ is defined by the spectral difference approximation T. The spectrum difference approximation T defines the difference between the spectrum of each frame and the spectrum of the receiving frame for each frame. Obviously there are many definitions that can be used for this kind of spectral difference divisor. The definition in the current preferred embodiment is as follows: Γ = log τγ + Σ Lp (N)

\ V hpj) NE is the RMS energy of the current frame, EP is the energy of the previous frame, L (N) is the Nth log zone ratio of the current frame, and LP (N) is the Nth of the previous frame Log zone ratio. The logarithmic area ratio L (N) is directly calculated using the Nth reflection coefficient kN, and the formula is as follows: 4Hickman200021tw; AND1P115.TW 84 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ---- ------- Installation -------- Order --------- ^ • 1! (Please read the precautions on the back before filling this page) 548631 A7 B7 Intellectual Property of the Ministry of Economic Affairs The erroneous transition error Ετ printed by the bureau's consumer cooperative is a function of the spectrum difference approximation 値 T, which is defined as follows: When both the previous frame and the previous frame are silent, or both are audible, Ετ is set to = 〇; otherwise, Et = Gt + At / T, where T is the approximate spectral difference of the current frame. The definition of the pronunciation change error can also vary greatly. The main characteristic of the pronunciation transition error defined here is that as long as the pronunciation state changes (voiced to silent, silent to voiced), the loss is estimated, which is a decreasing function of the spectral difference between the two frames. That is, unless significant spectral changes occur at the same time, changes in pronunciation state are not welcome. Such a pronunciation transition misdefinition can reduce the processing time required to provide a good pronunciation state decision, and provides a considerable advantage in the present invention. Other errors Es and EP that constitute a transition error in the presently preferred embodiment can also be defined variously. That is, the utterance status error can be defined in any way, as long as it matches the data in the current frame with a good tone period and overwrites the data with a bad match during the period. Similarly, the pitch deviation error EP, which is consistent with the change during the pitch, can be defined in any way. Although preparation for doubling and halving is best, tonal deviation errors need not include preparation for doubling and halving. Another optional feature of the present invention is that if the tonal deviation error includes preparations for doubling and halving of tracking, it is best to find the best curve. 4Hickman200021tw; AND1P115.TW 85 This paper size applies Chinese national standards ( CNS) A4 specification (210 X 297 mm) (Please read the notes on the back before filling out this page) 装 · 11111111 #. Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Description of Invention (§ς) Double (or halve) the tone period along the best curve to make it as consistent as possible. It is also important to note that the three components identified from the transformation errors do not necessarily need to be used in their entirety. For example, if "goodness" 过滤 lower tonal hypothesis is filtered out in the previous stage, or "goodness" 値 is used to rank the tone periods, and when goodness 値 higher tonal periods are selected, the pronunciation state errors can be omitted. Similarly, other components can be included in the transition error definition as needed. In addition, please note that the dynamic programming method taught by the present invention does not need to be applied to the tone period candidate 摘 extracted from the filtered residual signal after adjustment, nor does it need to be applied to the tone period candidate 衍生 derived from the LPC residual signal, but it may be applied. Apply to any tone period candidate, including tone period candidate directly extracted from the original input voice signal. These three errors are then added together to provide the total error between a tone candidate in the current frame and a tone candidate in the previous frame. As mentioned earlier, these transition errors must be accumulated and added to provide the cumulative loss of each curve in the dynamic programming calculation. This dynamic programming method, which finds both tones and pronunciation at the same time, is peculiar in itself and is only used in conjunction with this method of finding candidate cymbals during tones. Any method to find candidate cymbals during tones can be combined with this peculiar dynamic programming 4Hickman200021tw; AND1P115.TW 86 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) '" " " ^^ Packing -------- Order --------- (Please read the notes on the back before filling this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Invention Explanation (M) calculus is used in combination. Regardless of which method is used to find candidate 値 during tone, the candidate 値 is simply provided as input to a dynamic programming algorithm. The preferred embodiment of the present invention uses a small computer and high precision sampling, so this system is not economical for large applications. Therefore, a better mode for implementing the present invention in the future should be a specific embodiment using a microcomputer-based system, such as a Ti professional computer. This professional computer is equipped with a microphone, a speaker, and a voice processing card including a TMS 320 digital processing microprocessor and a data converter, which is enough hardware to implement the present invention. Voice Access Authentication for Data Access Figure 21 illustrates a specific embodiment of the present invention that uses speech recognition to allow users to access data on the network. When a user requests access to data, such as a website, operation 2100 prompts the user to provide a voice sample. In job 2102, a voice sample of a user is received through a network. At operation 2104 ', retrieve the user's registration information. It is important to note that the information may be retrieved from a local storage device or from a network. The registration information includes a voice scan of the user's voice. After the user's voice sample is obtained, the voice sample and the voice scan in the registration information are compared in operation 2106 to verify the user's identity. Assignment 2106 is discussed in detail later. After verifying the identity of the user in operation 2106, the user data access permission will be granted in operation 2108. If, as a result of the verification of operation 2106, the user's identity is incorrect, access to the data will be denied in operation 2110. 4Hickman200021tw: AND IP 115.TW 87 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) ----- 11 --- · 11 ----- Order ----- 丨-丨 丨 -_ (Please read the notes on the back before filling this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Description of Invention (g?) This specific embodiment is particularly practical in the field of e-commerce. Third parties can be exempted from certification and certification. For more information on the procedures and devices for performing the above operations, read the instructions below and refer to Figures 22-27 and 29-34. In a specific embodiment of the present invention, a user's voice is recorded to establish a voice scan, and the voice scan is stored. This can be part of the registration process. For example, when the user is prompted to provide voice during the registration process, the user can speak into a microphone connected to his computer, and the generated voice data will be transmitted to a website for storage via the Internet, such as the Internet Capture of Verification Procedure. Then, when the user wants to access the website or a specific part of the website, he will be prompted to provide a voice sample, and this sample will be received and compared with the voice data stored in the website. Alternatively, the voice scan can include the user's password. Preferably, the voice scan includes multiple words spoken by the user to increase security. In such a specific embodiment, the voice scan can, for example, store multiple sets of passwords, and the user must save all of the passwords. The password provides voice samples. Alternatively, different access levels or different parts of the data may require different words. You can also use different words as browsing controls. For example, when you connect words to a specific page on the website, the user will be prompted to enter a password, and then display the website page related to the password according to the received password. Allowing several words to be included in the voice scanning at noon also allows identity verification by comparing and replacing words. For example, if the first word cannot identify the user's body 4Hickman200021tw; AND1P115.TW 〇〇 This paper standard applies Chinese National Standard (CNS) A4 size (210 X 297 mm) Packing -------- Order --------- (Please read the notes on the back before filling this page) System 548631 A7 B7 5. The description of the invention (q) can prompt the user to recite other words. For example, the user's voice sample and voice scan are very close, but when the difference between the two is above a predetermined threshold, the user can be asked to recite another word, which can be used to verify the user's identity. This allows the user multiple opportunities to access the data, and this feature is also particularly useful in the event that the user's voice changes slightly due to a cold, etc. The user's voice samples and / or the time and date of receiving the voice samples can be selectively recorded. Referring to operation 2106 of FIG. 21, an exemplary embodiment of the present invention is a system and method for establishing a positive or negative identity of a speaker, in which at least two different voice authentication devices are used, and can be used to supervise access to a security system Regulatory authority. In particular, the present invention can be used to provide voice authentication with very low false acceptance and false rejection rates. The "security system" referred to here is any website, system, device, etc. that only authorized individuals are allowed to access or use. Every time a user wants to access or use the system or device, he must be authenticated or identified with a positive authorization. . The principles and operations of the voice authentication system and method implemented in accordance with the present invention can be better understood with reference to the drawings and accompanying explanations. Next, referring to the drawings, FIG. 22 illustrates the basic concept of the voice authentication system for controlling access to the security system. 4Hickman200021tw; AND1P115.TW 89 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) -------- Order --------- (Please read the note on the back first Please fill in this page again for matters) 548631 A7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs B7 V. Invention Description (^ 0) The speaker 2220 communicates with the security system 2222 and the security center 2224 simultaneously or sequentially. The voice of the speaker 2220 is analyzed by the security center 2224 for verification. If the security center 2224 establishes a positive authentication, it will send a communication instruction to the security system 2222, and establish the correct identification code (ID) of the speaker 2220 as shown in 2226. And allows the speaker 2220 to access the security system 2222. The prior art system of FIG. 22 uses a single voice authentication calculus system. Therefore, this system will encounter the problem that the false acceptance and false rejection rates increase and decrease each other, resulting in a too high false acceptance rate and / or false rejection rate. Respectively leading to insecure and / or inefficient systems. The present invention is a system and method for establishing speaker identity using at least two different speech authentication calculation systems. Choosing two very different speech authentication calculus systems (such as text-dependent and text-independent calculus systems) can ensure that the two are not completely correlated statistically in terms of false acceptance and false rejection events, which is r < 1.0, where "r" is the statistical correlation coefficient. Assume that these two types of speech authentication algorithms are completely independent (r = 0), and the error rejection threshold of each algorithm is set to a low value, for example, 0.5%. Then, according to the rules of selection, and J. Guavain, L. Lamel and; b Prouts (March, 1995) The prediction in Figure 1 of the LIMSI 1995 scientific report. The error acceptance rate of each calculus system should be very high, in this case about 8%. 4Hickman200021tw; AND1P115.TW ΟΛ This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) ----------- installation -------- order ---- -(Please read the notes on the back before filling this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Description of the invention (ί (() However, if the speaker is positive in both calculation systems A positive identity is established only in the case of attestation. The joint false acceptance rate of the two should be (8% _ 2), which is 0.6%, and the joint false rejection rate should be 0.5% χ 2, which is 1%. As the degree of correlation between sets of calculus systems increases, the expected rate of joint false acceptance should increase, while the rate of false rejection should decrease. When fully correlated (r = 1.0), the combined rate of the two in this example will not increase. Reset to 0.5% and 8%. Please note that the best EER 値 for the calculation system characteristics used by B · Prouts is 3.5%. Extrapolate the graph of B · Prouts to a 2% EER 値 calculation system (currently the latest Technology), you can choose to set the false rejection rate at 0.3%, and then the false acceptance rate falls to 4.6% Right, to get a joint error acceptance rate of 0.2% and a joint error rejection rate of 0.6%. Therefore, the "different calculation system" used in this description and the scope of the patent application later refers to a calculation system with correlation r < 1.0. Please refer to FIG. 23, which shows a system for establishing speaker identity according to the present invention, hereinafter referred to as system 2350. The system 2350 includes a computerized system 2352, which includes at least two voice authentication calculation systems 2354. , Marked as 2354a and 2354b respectively. 4Hickman200021tw; AND1P115.TW 91 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm)

Aw -------- ^ --------- (Please read the notes on the back before filling out this page) Employees ’Consumption of Intellectual Property, Ministry of Economic Affairs 548631 A7 _____B7__ V. Description of Invention (Gas 1) The calculation system 2354 is selected to be different from each other. They are respectively responsible for independently analyzing the speaker's voice and obtaining independent positive or negative voice authentication. If each calculus system 2354 provides a positive attestation, it will identify the speaker as the correct user, but if one of the 2354 calculus systems generates a negative attestation, it will identify that the speaker is not the correct user (that is, it is identified as impersonation) By). Both text-related and text-independent speech authentication calculation systems can be used. Application examples include those described in the following U.S. patents: first extracting features of U.S. Patent No. 5,666,466, and then performing graphic matching calculations; neural network voice verification of U.S. Patent No. 5,461,697; U.S. Patent No. 5,625,747 Dynamic Time Warping (DTW) calculus; Hidden Markov Mode (HM%) calculus of US Patent No. 5,526,465; and Vector Quantitative (VQ) of U.S. Patent No. 5,640,490 Calculus. All references cited above are incorporated herein. According to a preferred embodiment of the present invention, the error rejection thresholds of the two calculus systems of 2354 are both set to a level of 0.5% or lower, which can be lower than or equal to 0.4%, and preferably lower than 0.3%. It is best if it is less than or equal to 0.2% or approximately equal to 0.1%. According to the actual application situation, the speaker's voice can be directly accepted by the system 2352, or by the remote communication mode by the system 2352. 4HICKMAN / 200021TW; AND1P115.TW 92 This paper size is applicable to Chinese National Standard (CNS) A4 (210 X 297 mm) (Please read the note on the back before filling in this page) Installation --- r --- Order --------- Attack 548631 A7 B7 V. Description of the Invention (e) Therefore, according to the preferred embodiment, the speaker's voice is accepted by the computerized system 2352 through the remote communication mode 2356 for analysis. For example, the remote communication mode 2356 may be a general or mobile telephone communication mode, a computer telephone communication mode (such as the Internet or an intranet), or a wireless communication mode. In Figure 23, these communication modes are represented by a unified telephone symbol. These communication modes will communicate with at least one receiver 2358 (two 2358 in the picture, 2358a and 2358b) in the computerized system 2352, as shown in the figure. Shown as a polyline. According to another preferred embodiment of the present invention, the computerized system 2352 includes at least two sets of hardware installations 2360 (there are two groups in the figure, 2360a and 2360b), and each set of hardware installations 2360 is responsible for making a set of voice Forensic calculation system 2354 works. The hardware installation 2360 may be of any type, including, but not limited to, a personal computer (PC) platform or equivalent device, a dedicated interface card in the computer, and the like. Different hardware installation 2360 can be installed at the remote end. "Remote" here refers to a situation where different installations 2360 communicate via remote communication media. Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs ------------ Crack— (Please read the precautions on the back before filling out this page) Ritual · In an application of the present invention, at least in safety One set of hardware installations 2360, such as 2360a, is implemented on system 2362, while another set of hardware installations 2360, such as 2360b, is implemented at least on Security Center 2364. In a preferred specific hardware installation, the hardware installation 2360b performed on the security center 2364 will be the same as the hardware installation 2360a performed on the security system 2362. 4HICKMAN / 200021TW: AND1P115.TW 93 This paper is applicable to China Standard (CNS) A4 specification (210 X 297 mm) A7 548631 B7 _ V. Invention description (called) information, and finally establish the positive or negative identification of the speaker on the security system 2362, and the specifications in the scope of patent application later The term "security center" is used to refer to a computer system that is responsible for performing at least one voice authentication algorithm system, and for performing positive or negative speaker identification procedures. According to a preferred embodiment of the present invention, the computerized system 2352 further includes a speech recognition calculation system 2366. The calculus system 2366 is responsible for identifying the spoken language changes of the speaker (closed to identify the speaker based on the speaker's speech) and operates the security system 2362 accordingly. The calculus system 2366 is further responsible for positively or negatively identifying spoken materials. If the spoken material and the identified speaker have a negative or positive association with the positive identity established through the calculus system 2354 as described above, the only relationship here is Only at the right time will the speaker be granted access to the security system 2366. The spoken materials spoken by the speaker may include any spoken words (at least one word), such as, but not limited to, name, ID number, and requirements. In a preferred embodiment of the present invention, a single security center 2364 with a voice authentication calculation system 54 is used to communicate with a number of security systems 2362, each of which has a different (second) voice authentication The calculation system 2354 allows the speaker to choose to access any of the security system 2362 subsets after passing the authentication. 4HICKMAN / 200021TW; AND1P115.TW οΔ This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) ----------- installation -------- order-- ------- (Please read the notes on the back before filling out this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Invention Description) Examples Please consider the following examples. The examples illustrate the present invention together with the foregoing description, but the scope of application of the present invention is not limited to this. Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs

El 24-27 are preferred embodiments of systems and methods not according to the present invention. Therefore, as shown in Figure 24, use its voice alone, or cooperate with a computer, wired phone, mobile phone, computer phone, transceiver (such as a radio transceiver) or any other remote communication medium, etc. connected to the network Device, user, such as speaker 2420, can communicate with security center 2424 and one or more security systems 2422, such as computer network (security system No. 1), voice mail system (security system No. 2), and / or a full array of Computer system (N security system) and other communications, of course, the communication range is not limited to the devices mentioned here. In a preferred embodiment, the speaker uses the telephone communication mode, and the phone numbers of all the security systems 2422 and the security center 2424 are the same, or the radio communication mode has the same frequency and modulation. In any case, the user should be able to communicate with the security system 2422 and the security center 2424 at the same time. In a preferred embodiment of the present invention, each security system 2422 includes only one receiver for authentication and authentication, and there is no transmitter. 4HICKMAN / 200021TW; AND1P115.TW 95 This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm). -------- Order ---- (Please read the precautions on the back before reading) (Fill in this page) 548631 A7 B7 V. Description of Invention (qt) Figure 25 illustrates the next step of the program. The Security Center 2424 analyzes the input speech using the following methods, (i) the prior art speech authentication 2530 calculus and (ii) the traditional spoken speech recognition calculus 2532, including, for example, the required speaker 2420 security system 2422 (1,2, ... or N number) Spoken ID of access code (formed at the same time), password and social security code. The false rejection threshold is set to a low value, for example, less than 0.5%, and preferably less than 0.3%, so that the false acceptance rate is maintained at about 4.6%. After the positive recognition of the input voice is established, the security center 2424 recognizes the speaking identity 2536 by transmitting a tone 2536 and the like. Speaker 2420 and a specific security system 2422 (e.g., based on the system access code used by speaker 2420) will receive a tone of 2536. Figure 26 is the next step. The security center 2424, or better security system 2422, uses a second voice authentication algorithm 2638 to perform authentication of the input voice. The second calculus is different from the voice authentication calculus 2532 used by the security center 2424 described earlier in conjunction with FIG. For example, the speech authentication calculation 2638 may be a neural network speech authentication calculation system as described in U.S. Patent No. 5,461,697. The false rejection threshold is also set to a low value, such as below 0.5%, preferably 0.3 or 0.1%. Therefore, according to the above reasoning and calculation, the calculation system of EER 値 is about 2%, and the degree of error acceptance (for example, 0.3%) is about 4.6%. 4HICKMAN / 200021TW: AND1P115.TW 96 This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm) (Please read the precautions on the back before filling this page) Installation II -------- -Printed by the Consumer Property Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 2422 are physically removed. As the identity processing in Security Center 2424 extends a pre-selected time interval, the simultaneous voice authentication in Security System 2422 will be t = .DELTA.T in the security system after receiving the tone 2536. 2422 starts. This time delay can ensure that the identification is not issued before receiving the approval of the security center 2422. As shown in Figure 27, the security system 2424 and the security system 2422 can only be established after the identifications 2742a and 2742b are established. Establish the final speaker identity 2740, giving the speaker access to the security system 2422. Therefore, only in the security center 2424 and the security system 2422 can a positive word be established After authentication, the speaker is confirmed, the completion procedure is confirmed, and access to the security system 2422 is allowed, as shown in 2744. If one of the systems 2422 and 2424 does not confirm the speaker's voice, the procedure does not count as confirmation completion, and its storage is rejected Take the security system 2422. Voice system used to manage the border. Figure 28 is a method to determine whether the user who passes the border is qualified based on the voice signal. First, at operation 2800, the voice signal of the user trying to cross the border is received. Then in Assignment 2802 analyzes the user's voice signal to determine whether the user meets the predetermined standards for crossing the border. Then in Assignment 2804 4HICKMAN / 200021TW; AND1P115.TW 97 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ) ----------- Installation -------- Order --------- (Please read the notes on the back before filling this page) 548631 A7 ___ B7 V. Description of the Invention (Journal) The output indication 'shows whether the user meets the predetermined border crossing standards. The detailed description of the procedures and devices for performing these operations is as follows. In the specific embodiment of the present invention shown in FIG. 28, Use voice signals to determine the identity of the user. The specific embodiment of the present invention can be used to allow an approved user to enter other countries across the border without presenting a written proof of identity. In such a specific embodiment, predetermined standards Can include identities contained in a list of people allowed to cross the border. For more information on procedures and devices for identifying people by voice, see "Voice Authentication for Data Access" earlier, and refer to Figure 22-27 earlier. And the method and apparatus mentioned below with reference to Figures 29 ^ 34. The negative voice signal is compared with many stored voice samples to determine the identity of the person. Each of these many speech samples is associated with the identity of a person. After comparing the voice signal and the voice sample, the identity of the person will be output after confirming the identity of the person. In addition to the identity of the person, the output can include a display for border guards, showing that the person is allowed to pass. Alternatively, where there are gates or turnstiles that prevent people from crossing the border or blocking entry into national borders, you can use this output to open the gates or turnstiles. In another embodiment of the present invention described in FIG. 28, the voice signal of a person can be used to detect emotions. Its predetermined criteria may include designing emotional basis criteria to assist in detecting smuggling and other illegal activities and to assist in capturing those who hold forged documents. For example, when detecting personnel answering questions from customs officials, 4HICKMAN / 200021TW; AND1P115.TW 9 < t This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) (Please read the notes on the back before filling in (This page) a Mamr II —Μ— I ϋ · 一 tfj0 ί 1 Met -ϋ ϋ I r Printed by the Consumers ’Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 Printed by the Consumer ’s Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of the Invention) Fear and anxiety in speech. Another type of emotion that can be detected is how tight the person is. For more information on detecting emotions in voice signals, see the previous chapters to understand how specific embodiments work. Figure 29 is a speaker identification method implemented by an application according to one aspect of the present invention. Assignment 2900 stores the scheduled first and last voice feature information store in the first location. Then input the voice data in the second place in homework 2902. Then, in step 2904, the voice data is processed at the second location to generate intermediate voice characteristic information. Assignment 2906 will transfer the intermediate speech characteristics information from the second location to the first location. Assignment 2908 will further process the intermediate speech characteristic information from the second location at the first location to generate the second final speech characteristic information. Assignment 2910 will determine at the first location whether the second and last speech characteristics information completely matches the first speech characteristics information, and generate a confirmation judgment signal indication. Figure 30 illustrates a method for identifying a speaker according to a second aspect of the invention. Assignment 3000 will process many sets of first and last speech feature information at the first location and store the corresponding identification information. Assignment 3002 will enter one of the voice data and identification information at the second location. Assignment 3004 transmits an identification to the first location. Assignment 3006 transmits one of the first and last speech characteristics information and the judgment coefficient corresponding to one identification information to the second location. The speech is then processed at a second location in job 3008 to generate a second final speech characteristic information. Assignment 3010 will determine whether the second final voice characteristics information is the same as the 4th HICKMAN / 200021TW; AND1P115.TW 99 at the second location according to the judgment coefficient. -------- Order --------- (Please read the notes on the back before filling this page) A7

548631 B7 V. Description of the invention ((0〇) A final speech characteristic information completely matches, and a judgment signal indicating confirmation is generated. According to the third aspect of the present invention, the speaker recognition system includes: processing speech data to generate standard speech characteristics A registration unit that stores and stores standard voice characteristic information; a first processing unit that inputs test voice data and processes the test voice data to generate intermediate test voice characteristic information; and is in communication with the first processing unit to receive intermediate test voice characteristic information, A second processing unit that further processes the intermediate test voice characteristic information to generate test voice characteristic information is connected to the registration processing unit to determine whether the test voice characteristic information completely matches the standard voice characteristic information. According to a fourth aspect of the present invention Application, the speaker recognition system includes a first processing unit that processes speech data to generate standard speech characteristic information and stores standard speech characteristic information and related ID information; selectively connects to the first processing unit to input related ID information and test The second processing unit of the audio data transmits the relevant ID information to the first processing unit of the first processing unit, generates test voice characteristic information according to the test voice data, and determines whether the standard voice characteristic information basically matches the test voice characteristic information. Two processing units. Next, please refer to the drawings, especially FIG. 31, to explain the basic elements of speaker recognition. The user speaks into the microphone 3101 to input his voice. The voice regular sampling unit 3103 samples the voice input data at a predetermined frequency. 4HICKMAN / 200021TW; AND1P115.TW 100 This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm). -------- Order --------- · (please first Read the notes on the back and fill in this page) Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Description of the invention ((〇 \) Voice characteristics information extraction unit 31〇 4 will extract the predetermined speech characteristic information or the final speech characteristic graph of each sampled speech data set. The above input and extraction process is performed for registration or initial processing After that, the mode selection switch 3108 will be closed to connect to the registration unit 3106, and the voice characteristic information is stored in the speaker identification information storage unit 3105 as the speaker's standard voice characteristic information. At the same time, the speaker identification information is stored. Then refer to FIG. This figure illustrates the information stored in the speaker identification information storage unit 3105. The speaker identification information includes the speaker's name, identification number, date of birth, social security number, etc. The stored information includes the speaker mentioned above. Standard speech characteristics information of the speaker corresponding to the identification information. As mentioned earlier, the standard speech characteristics information is generated by the speech processing units 3103 and 3104, which extracts the speech characteristics from the predetermined speech data entered by the speaker in the registration process. Graphics. The final speech characteristic information or speech characteristic graph contains a series of the above-mentioned speech parameters. Returning to FIG. 31 again, when the mode selection switch is turned off to connect the speaker recognition unit 3107, the speaker recognition process is executed. The user must first enter his or her speaker identification information, such as a number, through the identification input device 3102 to be identified as a registered speaker. The registration device 3106 designates the corresponding standard speech characteristic information or the final speech characteristic pattern stored in the speaker identification information storage unit 3105 according to the identification information, and transmits it to the speaker identification unit 3107. The user also speaks some predetermined words through the microphone 3101 to input his voice data. The input voice data is determined by the voice 4HICKMAN / 200021TW; AND1P115.TW 1〇1 This paper size is applicable to China National Standard (CNS) A4 specification (210 X 297 mm). -------- Order ---- ----- (Please read the precautions on the back before filling out this page) Printed clothing by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Description of the invention ((c) period sampling device 3103 and speech characteristic parameter summary Take the device 3104 processing to generate test voice characteristic information. The speaker recognition unit 3107 compares the test voice characteristic information with the standard voice characteristic information specified above to determine whether the two completely match. The speaker recognition device 3107 then According to the above comparison, a judgment signal indicating the above-mentioned substantially consistent state is generated. The above and other elements of the speaker recognition concept are all implemented on a computer or a telephone network according to the present invention. The computer-network type speaker recognition system should have many copies. End processing units, and at least one management processing unit. The network should also share a common database that is usually located in a central management processing unit. In general, The computer-networked speaker recognition system is distributed at both ends of the spectrum. One end of the spectrum is characterized by a large amount of voice input processing at the local end, and the other end is characterized by a large number of voice input central processing. In other words, to complete the speech In the case of speech recognition, the local input processing unit, the central processing unit, or a combination of the two are used to process the speech input to determine whether it is roughly consistent with the registered speech data specified previously. However, the computer network used in the present invention does not have to be It is limited to the above-mentioned central-to-terminal restrictions, and may also include other systems such as a decentralized system. Next, please refer to FIG. 33 to describe a specific embodiment of the speaker recognition system according to the present invention. The local processing units 3331-1 to 3331-11 respectively use The network lines 3333-1 to 3333 · η are connected to the management central processing unit 2332. The local processing units 3331-1 to 3331-η all include a microphone, a voice periodic sampling unit 3103, and a voice characteristic parameter extraction unit 4HICKMAN / 200021TW; AND1P115.TW 102 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) -------- Order ------ (Please read the notes on the back before filling out this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 a? B7 V. Description of Invention (⑹) 3104 And a speaker recognition unit 3107. Each of the local processing units 3331-1 to 3331-11 can input voice data and process voice input to determine whether its characteristic pattern roughly matches the corresponding standard speech characteristic pattern. Management Central The processing unit 3332 includes a speaker identification data management unit 3310, which can perform management functions including registering and updating standard voice characteristic information. Next, referring to FIG. 34, a specific embodiment of the speaker recognition system described above will be further described. To simplify the explanation, we only detail the remaining components of a local processing unit 3331-1. In order for the local processing unit 3331-1 to communicate with the management processing unit 3332 through the communication line 3333-1, the local processing unit 3334-1 provides a first communication input / output (I / O) interface unit 3334-1. Similarly, the management processing unit 3332 includes a second communication I / O interface unit 3333-1 at the other end of the communication line 3333_1. In the following, the registration and identification procedure will be roughly explained using the preferred embodiments described earlier. When the user registers the standard voice characteristic information, the user needs to speak a predetermined set of characters through the microphone 3101 to input voice data, and input the user identification code through the ID input device 3102. The mode switch 3108 is placed in the registration mode, and transmits the processed voice characteristic information to the registration unit 3106 through the interfaces 3334-1, 3435 and the communication line 3333-1. The registration unit 3106 controls the speaker registration information storage unit 3105 to store the voice characteristic information and the speaker identification code. 4HICKMAN / 200021TW; AND1P115.TW 103 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) ----------- installation -------- order-- ------- (Please read the notes on the back before filling out this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 Α7 V. Description of the invention ((οψ The user specifies it through the user ID input device 3102 ID information for future speaker identification. The input information will be transmitted to the management processing unit 3333 · 1 through the interfaces 3334-1, 3435 and communication lines 3333-1. The management processing unit 3332 will take a response and send The standard voice characteristic information corresponding to the specified user ID is transmitted to the speaker recognition unit 3107. The selection mode switch is set to the speaker recognition mode to connect to the speaker recognition unit 3107. The user also uses the microphone 3101 to input his voice input and samples periodically The unit 3103 and the voice characteristic information extraction unit 3104 then process the voice input to generate test voice characteristic information and output it to the speaker recognition unit 3107. Finally, the speaker recognition unit 3107 judges the measurement Whether the voice characteristic information is roughly consistent with the selected standard voice characteristic information. The output judgment signal will show this judgment, authorizing the local processing unit 3331 · 1 to continue to perform more changes related to the management processing unit 3332. In summary, the above comparison The preferred embodiment basically processes the input voice data on the local processing unit. Using voice suppression and browsing on the Internet Figure 35 illustrates a method for identifying voice commands for operating data on the Internet. First, in homework 3500, provide data on the website. In homework 3502, a voice signal is received from a user who accesses the website. In homework 3504, the voice signal is interpreted to determine a browsing instruction. In homework 3506, The selected information of the website is output according to the browsing instructions. 4HICKMAN / 200021TW: AND1P115.TW 1〇4 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm). ------- (Please read the notes on the back before filling this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 V. Description of the invention (丨5) In the specific embodiment of the present invention, the data includes a voice-activated application. In this specific embodiment, the browsing instruction can control the execution of the application. In an embodiment of the application of the present invention, the voice signal can be transmitted. Carry out Internet financial practice. Users can access websites from computers, phones, or both. They can choose to output selected data to phones. Such specific embodiments can be used for mail services. For example, you can use text-to-speech technology to “write” an e-mail over the phone without using a display. You can also use text-to-speech technology to “read” email over the phone. You can judge the language based on the voice signal. Then use the language spoken by the user to interpret the voice signal to determine the instruction. This is particularly useful on international customer service systems on the Internet. There is an option to use artificial intelligence to interact with the user, including verbal responses and the like. Contents and Applications of Spiritual Sound Control FIG. 36 is a general block diagram of an information system 3610 for controlling content and applications via a network using a voice signal according to an embodiment of the present invention. The information system 3610 includes an information distribution center 3612. This information center receives information from one or more remote information providers 3614-1, ..., 36U-n, and supplies or broadcasts this information to the terminal unit 3616. This 4HICKMAN / 200021TW; AND1P115.TW 1〇5 This paper size is applicable to China National Standard (CNS) A4 specification (210 X 297 mm) Packing -------- Order --------- ( (Please read the precautions on the back before filling this page) 548631 A7 Printed by the Consumer Property Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs B7 5. Information used in the description of invention (丨 Qin) includes but is not limited to analog video, analog audio, digital Text services such as video, digital audio, news reports, sports scores, stock market quotes and weather reports, electronic messages, electronic program guides, database information, software such as game programs, and wide area network data. The information distribution center 3612 can also choose to choose another location or generate the information locally, and provide the terminal unit 3616 with this locally generated information. The information transmitted by the information distribution center 3612 to the terminal unit 3616 includes vocabulary representations of spoken sounds or words (pronounced) words. This vocabulary provides, for example, spoken control of device 3618 and spoken control of access to information transmitted by information distribution center 3612. In particular, the terminal unit 3616 receives vocabulary data from the information distribution center 3612 and speech (pronunciation) data from the user. The terminal unit 3616 includes a processor that executes a speech recognition algorithm to compare vocabulary data with spoken command data to recognize, for example, a command from the control device 3618 or a command to access information transmitted by the information distribution center 3612. The terminal unit 3616 suitably generates a command to control the device 3618 or to access the information transmitted by the information distribution center 3612. The speech recognition calculus used here refers to the calculus that converts spoken voice input into text or corresponding instructions. The speaker authentication algorithm refers to an algorithm that verifies the required speaker identity based on the voice samples of the requester. The speaker recognition algorithm refers to the algorithm for identifying a speaker from a previously sampled selection list based on the speaker's voice input. Speaker recognition algorithms can be used, for example, to limit a particular dictator's ability to control devices and / or access information. 4HICKMAN / 200021TW; AND1P115.TW 106 ----------- install -------- order --------- (Please read the precautions on the back before filling this page ) This paper size applies to China National Standard (CNS) A4 (210 X 297 mm)

548631 Printed by A7 B7, Consumer Cooperatives, Intellectual Property Bureau, Ministry of Economic Affairs 5. Description of the invention ((.) The information transmitted from the information distribution center 3612 to the terminal unit 3616 may be, for example, phoneme data. Phoneme is capable of distinguishing languages Or the smallest unit of sound in different pronunciations in dialects. Therefore, each sound or spoken text in a vocabulary can be represented by a group of phonemes. In addition, vocabulary data can be template data generated by a person or a group of people reading each sound or word. Each spoken sound or word in the vocabulary can be represented by a respective corresponding template. It should be noted that although the system of FIG. 36 illustrates a system in which the information from the information providers 3614-1, ..., 3614-n Information and vocabulary data are transmitted through the same communication link, but the present invention is not limited to such applications. Therefore, information and vocabulary data from information service providers 3614-1, ..., 3614-n can be transmitted through different Communication link for transmission. Voice data can be provided to the terminal unit 3616 using many different arrangements. In the first example In a non-limiting arrangement, a remote control is provided, including a wireless microphone or related converter, to transmit the voice or words spoken by the user to the terminal unit 3616 via electrical, optical or wireless frequency signals. Terminal The machine unit 3616 includes a receiver, an analog front end that adjusts the received signal, a codec that performs analog to digital conversion of the modulated signal, and an interface circuit connected to the processor. The adjustment operation is Refers to the modification of electrical signals received from a voice converter by noise reduction, noise reduction, filtering, and other known techniques. In the second exemplary but non-limiting arrangement, the use of a microphone Analog receiver for adjusting the microphone sound signal, stomach for performing the adjusted signal

4HICKMAN / 200021TW; AND1P115.TW This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm) 1 ^^ · Packing -------- Order --------- Awr (Please read the precautions on the back before filling out this page) 548631 Printed by the Consumer Property Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Invention Description ((ο Poor) Codec with ratio to number k conversion, and use, for example, The infrared or radio signal transmits the digital audio data signal to the transmitter of the terminal unit 3616 for remote control. The terminal unit 3616 includes a receiver that receives the digital audio data signal and an interface circuit connected to the processor. Digital audio Data signals usually require a transfer rate of at least 64 kbits per second. In a third exemplary but non-limiting arrangement, a microphone, an analog receiver for § weekly microphone sound signals, and analog to digital signals after adjustments are performed A converted codec, a digital signal processor that analyzes digitized sound signals to extract spectrum data, and uses, for example, infrared signals to transmit spectrum data to the terminal unit 3616 Transmitter for remote control. The terminal unit 3616 includes a receiver for receiving spectrum data and an interface circuit connected to the processor. Since the third arrangement transmits the spectrum data 'and the digitized sound data in the second arrangement The data transmission rate is very low, which is lower than 3610 kbits per second. Because the spectrum analysis is performed in remote control, compared with the second arrangement, during the identification operation, the terminal unit 3616 The processor load is reduced by about 30-50%. In the fourth exemplary arrangement, the terminal unit 3616 has a microphone, an analog front-end for adjusting the microphone sound signal, a codec that performs analog-to-digital conversion of the adjusted signal, And the interface circuit connected to the processor. In the fifth exemplary arrangement, the terminal unit 3616 has a microphone, an analog front-end for adjusting the microphone signal, a codec that performs analog-to-digital conversion of the adjusted signal, and analyzes the digitized signal A digital signal processor to extract spectrum data, and an interface circuit connected to the processor bus. The fourth arrangement, the digital signal in the fifth arrangement 4HICKMAN / 200021TW: AND1P115.TW 108 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) --------- --AW ^ -------- Order --------- (Please read the notes on the back before filling out this page) 548631 A7 _ B7 V. Description of Invention (i〇Cj) Processor It is used to reduce the load of the processor of the terminal unit 3616. The above-mentioned various arrangements are merely examples, and other arrangements can be used to provide the terminal unit 3616 with voice data within the scope of the present invention. The vocabulary data transmitted by the information distribution center can define the instructions that the user speaks to the control device 3618. The device 3618 may be any device capable of operating in response to instructions provided by the user, and the present invention is not limited in this regard. Thus, the device 3618 may be, for example, a television, a stereo radio, a video player, a tape recorder, a CD player, a video player, a video game instrument, or a computer. In terms of holding, it is assumed that the device 3618 is a computer whose power source is connected to the terminal unit 3616 that can switch on and off the power socket, and as long as the user speaks the commands of "power on" and "power off", the computer can be turned on and off. The information distribution center 3612 then transmits to the terminal unit 3616 the phoneme or template vocabulary data that defines the command words including the words "on" and "off". When the user says "turn on" or "turn off", and provides the voice data of the corresponding instruction to the terminal unit 3616 by using any of the above arrangements, the processor of the terminal unit 3616 will execute a speech recognition algorithm To compare spoken instructions with phoneme or pattern data representing the vocabulary of the instructions to identify spoken instructions. The terminal unit 3616 controls the device 3618 appropriately, that is, turns on or off the power of the computer. Since the computer is plugged into the switchable power socket of the terminal unit 3616 as described above, the power on and off actions of the computer are performed inside the terminal unit 3616. However, the present invention is also applicable to transmitting the identification command to the device 3618 for communication via 4HICKMAN / 200021TW; AND1P115.TW 109 This paper size is applicable to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) (Please read the back (Please fill in this page again)

II n ϋ .ϊ, a— ϋ ϋ -ϋ ϋ ^ 1 II Printed by the Employees 'Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Printed by 548631 A7 B7 V. Implementation of the invention (U0) situation. For communication purposes, this type of communication link may be the Internet, infrared link, RF link, coaxial cable, telephone network, satellite system, or optical fiber, and the present invention is not limited to these types. The vocabulary data may also select or otherwise define the words and instructions that the user said when accessing the information transmitted by the information distribution center 3612. This feature allows users to perform tasks that are difficult to perform with a user interface that operates in a menu mode. For example, you can use this feature to use the "Search Key" command to perform a keyword search on news report titles from the information distribution center 3612. In particular, the information distribution center 3612 will determine which individual words are keywords and generate These keywords map to the phoneme or template phoneme or template "dictionary." The information distribution center 3612 transmits the news reports and the dictionary to the terminal unit 3616, and the terminal unit stores the data in the memory. For each keyword, the terminal unit 3616 uses the dictionary to generate a corresponding phoneme or template string. The string then uses the speech recognition algorithm "register" as a single recognizable pronunciation, that is, it will become a basic part of the speech recognition algorithm vocabulary. Registration includes assigning identifiers to phonemes or pattern strings, which can be numbers or keywords themselves. When the user then says the "search keyword" command, the display connected to the terminal unit 3616 or the computer connected to the terminal unit 3616 will display a dedicated screen for this command. The user can say "only the key" "Word" command to limit the scope of the news reports from the terminal unit 3616 search information distribution center 3612 to the news with the keyword in the title. Then the user can say other keywords to indicate a more detailed search , Or view the target

4HICKMAN / 200021TW; AND1P115.TW 11C This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm) Surface — ----------- ^ Packing -------- Order --------- ^ 9 (Please read the notes on the back before filling out this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A / B7 V. Description of the invention News coverage of said keywords. It is clear that it is very difficult to perform such operations using the traditional menu-based user interface. Figures 37A, 37B, and 37C are block diagrams of a consumer television system employing the present invention. Of course, the present invention can also be applied to other information systems other than the user television system, and the application of the present invention is not limited to this. The customer television system provides information to a number of customer locations, such as 3720-1, ..., 3720-n (refer to Figure 37C). Information may include, but is not limited to, analog video, analog audio, digital video, digital audio, news reports, sports scores, text services such as stock prices and weather reports, electronic messages, electronic program guides, database information, software such as games, and wide area networks Information, etc., and not limited to these. Referring to FIG. 37A, the user television system includes a number of information providers 3714-1,..., 3714-n, each of which may provide one or more of the aforementioned information formats. For example, the information provider 3714-2 includes an information source 3715 that provides an analog television signal to the transmitter 3718. Transmitter 3718 is connected to the Internet uplink 3721, which transmits analog television signal 3722-2. The information providers 3714-1 and 3714-3 provide digital information from the information source 3715 to each encoder 3716, which generates an encoded data string for transmission. Information sources 3715 of information providers 3714-1 and 3714-3 may be memory used to store information, such as optical memory. If any of the information providers 3714-1 and 3714-3 provides various information, such as many different game programs, or different types of text services, or perhaps most television or audio programs, the encoder 3716 will multiplex the information to generate Multiplexed data strings for transmission. Encoder 3M6 4HICKMAN / 200021TW; AND1P115.TW 111 ----------- install -------- order --------- (Please read the note on the back first Please fill in this page for 5 items) This paper size applies to China National Standard (CNS) A4 (210 X 297 mm)

548631 A7 Printed clothing by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs Β7 V. Description of invention (\ H) The material string will be provided to the transmitter 3718 and then to the Internet uplink 3721. Looking at the example of Fig. 37A, the encoder 3716 operated by the information provider 3714-1 generates a digital data signal 3722-1, and the encoder 3716 operated by the information provider 3714-3 generates a digital signal 3722 -3. Each of the signals 3722-1, 3722-2, and 3722-3 is transmitted to the original point installation 3725 via the Internet 3723 (refer to FIG. 37B). We know that there will be many information providers in the system of the present invention, and therefore many signals can be transmitted to the Internet through 3723, such as the location of the origin point installation 3725. Although not shown in the figure, the signal can be received at, for example, many locations other than the origin point installation, such as the location of a Direct Broadcast Service (DBS) user. In addition, although the connection between the information provider and the origin installation is represented by a network connection in the figure, the application of the present invention is not limited to this. Thus, this link may be, for example, a coaxial cable, a telephone network, a satellite system, the Internet, a radio frequency (RF) link, optical fiber, or any combination of these media. In addition, although the distance between the information provider in FIG. 37A and the original point installation 37M is very long, the actual location of one or more information providers may be the same as the original point installation 3725. Referring to FIG. 37B, the downlink 3724 of the original point 3725 provides the received signals 3722-1, 3722-2, and 3722-3. The origin point installation 3725 is a communication hub that connects with various information providers and connects the information provider to the user's location 372 (M,..., 3720-η. For example, the received digital data The signal 3722-1 will be supplied to the receiver 3726-1, and then to the modulator 3728-1.

4HICKMAN / 200021TW; AND1P115.TW This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm) ----------- installation -------- order --- ------ (Please read the precautions on the back before filling out this page) 548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs V. Invention Description ((\,) Channel. Modulator 3728-1 may Use any appropriate modulation technique, such as Quadrature Partial Response (QPR) modulation. The received analog TV signal 3722-2 is sent to the receiver 3726-2, then sent to the encoder 3730 for encryption, and then to the modulator 3728-2, modulated to different cable channels. The encoder 3730 will also add in-band data to the analog TV signal 3722-2, which will be discussed in detail later. Obviously, the same can be done in this The remote or remote provides other receivers, modulators and optional encoders (not shown) for digital and analog information signals received from other providers. The received digital data signal 3722-3 is sent to Information Signal Processor (ISP) 3742 to transmit using so-called in-band or out-of-band transmissions. In order to send other providers' data strings (not shown) to the ISP 3742. The ISP 3742 is responsible for receiving one or more data signals and then transmitting the data to the user terminal location to be introduced soon. The ISP 3742 provides the data to the encoder 3730 ISP 3742 can also provide data to other encoders based on the amount of data to be transmitted and the speed of the data that must be provided and updated. Data is repeatedly sent by encoder 3730. If there is only one encoder, but the amount of data is large , The repetition rate will become very slow. Using multiple encoders can increase the repetition rate of the data. In particular, the encoder 3730 adds data to the frequency band for transmission to the user, and also related analog TV signals 3722- 2 plus. In one arrangement, the data will be placed in the vertical interval compression of the TV signal, not 4HICKMAN / 200021TW; AND1P115.TW 11. This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 mm) ( (Please read the notes on the back before filling out this page) ϋ · 1 ϋ I ϋ i.— I h-mouth, 1 I 11 I » Printed by the agency 548631 A7 B7 V. Description of the invention (\\ vp can also be placed in the signal, and the application of the invention is not limited to this. For example, it is well understood that data can be added using amplitude modulation In the sound carrier. The in-band transmission discussed here refers to the carrying of data in a video television channel that includes a carrier and a video carrier. Therefore, the data of ISP 3742 may be transmitted using the amplitude modulation of the sound carrier, hereinafter referred to as in-band audio data, or vertical or horizontal compression placed on analog TV signals, hereinafter referred to as in-band video data. It is also possible to arrange for the ISP 3742 to provide transmission of unused portions of digital data strings such as MPEG compressed video data strings. ISP 3742 can also receive and / or generate information locally. For example, ISP 3742 can generate messages such as event notifications or service interruptions or changes, and send them to users. After receiving the information from the data provider, the ISP 3742 can send it directly without modification, or reformat the format, and then provide it to the encoder 3730 for transmission to the user. ISP 3742 will also send information to origin point controller (" HEC ") 3732. HEC is connected to encoder 3730 and out-of-band transmitter 3734. Although HEC 3732 in the description is connected to the same encoder as ISP 3742 , But in fact HEC3732 can be connected to one or more other encoders. HEC 3732 can use Scientific-Atlanta Model 8658 to control the transmission of data to encoder 3730 and out-of-band transmitter 3734. As mentioned earlier, encoder 3730 will Put the data in the frequency band for transmission to the customer, together with the coding-related TV signals. The out-of-band transmitter 3734 is Lee 4HICKMAN / 200021TW; AND1P115.TW 114 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 (Mm) ------------ ^ -------- Order ------ (Please read the precautions on the back before filling this page) # · 548631 Ministry of Economy Wisdom Printed by A7 B7, Consumer Cooperatives of the Property Bureau. 5. Description of the Invention (\\ 5) Use another carrier, that is, do not transmit information in the same channel. In a specific embodiment, the 'out-of-band carrier is 108.2 MHz, but it can also be used. Other out-of-band carriers Information transmitted under the control of HEC 3732 can be, for example, encoded data. In one arrangement, information is inserted at each vertical compression interval to indicate the type of encoding used in the next video field. Encoding systems are now very common. For example, you can use synchronous compression coding, video inversion coding, etc., or a combination of different coding technologies. In addition, you can also send authorization information. Authorization information authorizes users to receive characteristic channels or programs. Received from ISP 3742 and / or HEC 3732 The received information can also be transmitted through data repeaters (not shown), such as the Scientific_ Atlanta Model 8556-100 data repeater, in the form of in-band audio or video data on non-encoded channels. Some of the transmitted information is comprehensive That is, it will be transmitted to every user. For example, the encoded data can be transmitted in its entirety. However, the fact that each user receives the encoded data does not mean that the terminal unit of each user can decode the received signal. Instead, only authorized user terminal units can decode the received signals. On the one hand, some information transmission is address transmission. For example, authorization information is usually addressed to individual users. That is to say, when transmitting, the data will have an associated address (such as the serial number of the user terminal unit). The terminal unit receives the information and responds according to the received information. The terminal units of other users ignore this information. In addition, the group addressing data can also be used, which affects the terminal units of the entire group of users. 4HICKMAN / 200021TW; AND1P115.TW 115 This paper size applies to China National Standard (CNS) A4 (210 X 297 public love). -------- Order --------- (Please read first Note on the back, please fill out this page again) 548631 A7 B7 Printed clothing by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs V. Invention Description (\ α) Modulators 3728-1, 3728-2, any other modulators, and out-of-band The output of the transmitter 3734 is sent to the combiner 3736. The combiner combines individual channels into a single broadband signal, and then sends it to many user locations 3720-1, ..., 3720-u through the distribution network 3738 (refer to Figure 37C). The distribution network 3738 may include one or more optical transmitters 3740, one or more optical receivers 3742, and a coaxial cable 3744. As shown in Figure 37B, a consumer television system can include many origin point installations, each of which provides information for many locations within a particular city or geographic area. The central control 37 can be used to coordinate the installation of each origin point in the user's television system. The Central Control 3746 is usually linked to the central office of a multi-service operator, and can communicate with control origins in many cities. The central control 3746 contains a system control computer 3748, which commands the other components of the central control 3746. An example of a system control computer 3748 is the Scientific-Atlanta System Manager 3610 network controller. Central control 3746 may, for example, provide billing services to service providers, including pay-per-view services. The billing computer 3750 stores billing information, and can also create billing formats and print bills. The system control computer 3748 and the HEC 3732 may communicate through a modem, although the present invention is not limited thereto. Authorization data can be transferred from the system control computer 3748 to HEC3732. HEC3732 then appropriately authorizes the format of the authorization data and transmits the authorization data to the user terminal unit in-band through the encoder 3730 or out-of-band through the out-of-band data transmitter 3734, as described above. 4HICKMAN / 200021TW: AND1P115.TW 116 This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm) ----------- installation -------- order-- ------- (Please read the precautions on the back before filling this page) 548631 Printed clothing A7 B7 of the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the invention (m) The original point installation 3725 also includes the user's location 3720-1, ..., 3720-n RF processor 37S2 receiving reverse path data. These data communications include pay-per-view viewing purchases forwarded to the system control computer 3748, as well as requirements for users to install the database information stored on the 3725 using the original point. For example, a database server 3754, such as the Oracle.RTM. Database server, provides users with access to reference materials, such as encyclopedias, atlases, dictionaries, and the like. The user request is forwarded from the RF processor 3752 to the information request processor 3756, which accesses the database 3754 for the requested information, and forwards the requested information to the requesting user, for example, through the aforementioned addressing band or outside the band Change transmission. In addition, the information request processor 3756 also accesses the communication network 3758 to provide users with access to other services, such as banking services. As the amount of data transferred between the origin point installation and the user's location increases, out-of-band and digital transfers can be used to increase usage. For example, a bandwidth of 50 MHz can be allocated for digital data (non-video) transmission of the forward channel (to the user terminal unit) and the reverse channel (from the user terminal unit). Allocate more than 200 MHz to digital video and 300 MHz to 500 analog video. Therefore, although many exemplary transmission technologies have been discussed previously, the present invention is not limited to the method of information communication between the origin point installation and the user's location. Referring to FIG. 37C, each user location 3720-1,..., 3720-n contains a user terminal network 3760 connected to a distribution network 3738. 4HICKMAN / 200021TW; AND1P115.TW 117 is used here. This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm). Order --------- (Please read the notes on the back before filling out this page) A7 548631 B7_ V. "User Position" of the invention description (丨 丨 8) refers to the 3725 remote end installed at the original point Anywhere. According to the present invention, the user terminal can be, for example, located in a home, a classroom, a restaurant room, a hospital, or an office. Each user terminal unit 3760 can be connected to one or more devices 3762-1, ..., 3762-n. The devices 3762-1, ..., 3762-n may include devices capable of responding to instructions given by the user, and the application of the present invention is not limited to this. Thus, the device may include televisions, stereo radios, video recorders (VCRs), recorders, CD players, video players, video game instruments, computers, and the like. Some specific devices can be connected and operated together. Therefore, as shown in Figure 37C, device 3762-1 is connected to device 3762-2. For example, device 3762-2 may be a television and device 3762-1 may be a video recorder. For the sake of explanation, it is assumed that the device 3762-1 is a video player and the device 3762-2 is a television. One or more of the devices 3762-1, ..., 3762-n can be connected to the switchable power socket of the user terminal unit 3760, whereby the user terminal unit 3760 can internally influence the opening or closing of these device switches. The remote control unit 3766 exchanges information with the user terminal unit 3760 through the communication link 3768. For example, the communication link 3768 may be an infrared link. Language Translation The system of the present invention uses a dictionary and a limited set of grammatical rules to translate a language. The dictionary contains four major types of language units. Each language unit is (1) a single word, such as "dog", or "government"; (2) a combination of words, 4HICKMAN / 200021TW: AND1P115.TW 118 This paper standard applies to Chinese national standards (CNS ) A4 size (210 X 297 mm) Packing -------- Order --------- (Please read the precautions on the back before filling out this page) Employee Consumer Cooperatives, Intellectual Property Bureau, Ministry of Economic Affairs Printed 548631 Printed clothing A7 B7 of the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 5. Description of the invention (丨 1) For example, π parking space '' or '' Prime Minister'1; or (3) Proprietary A name; or (4) a word with a unique definition in the present invention; or (5) one of multiple meanings with multiple interpretations. In the last case, each definition of a word represents a different linguistic unit, and each definition belongs to Items in different forms of categories. In order to facilitate automated processing, each definition will, for example, be distinguished by the number of periods that appear at the end of the suffix. There is no period after the list item representing the first (arbitrarily specified) definition, which represents the second After defining the listed items There will be a period, and so on. Or, you can use numbers, such as subscripts, to distinguish different word meanings. The words dedicated to the present invention occupy only a small part of the entire dictionary, and these words are not specific to the present invention and and It is based on different natural language word meanings. Instead, the specific word meanings of the present invention are relaxed to limit the total number of words in the dictionary. For example, in a preferred embodiment, the word meaning of "&use; use" is relaxed, which means that any Objects serve their primary purpose, so in the sentence "Jake use book", "use" stands for reading. " on "can be used to represent time (for example, (i go-to ballgame) on yesterday). However, if it is easier to use, you can cancel all the invention-specific words, and the dictionary will be expanded. The invention will allow the word The general dictionary is divided into four categories: "Things" or noun nouns, such as people, places, things, activities, or ideas' represented by T; "connective words", which specify two (or more) nouns (including the usual Words interpreted as prepositions and connectives, and words that describe relationships using actions, beings, or states of existence), represented by C; "descriptive words" 4HICKMAN / 200021TW; AND1P115.TW 119 This paper is applicable to this paper China National Standard (CNS) A4 Specification (210 X 297 mm) ----------- · Installation -------- Order --------- (Please read first Note on the back, please fill in this page again) 548631 A7 B7 Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economy Represented by D; and "logical connectives" to create a group Noun, represented by C. The preferred logical conjunctions are "and " (and) and" or ,, (or). The dictionary naturally cannot and will not contain a list of possible proper names; the proper names will be returned in angle brackets like other words not recognized by the present invention, indicating that there is no translation. The system also does not recognize verb tenses; the connectives are all expressed in the present tense, because tenses are easily understood from the context. However, a time, a specific day, and / or a date is still specified to represent the tense. According to the four expansion rules, sentences according to the present invention are constructed from nouns in the dictionary. The most basic sentence starts with the following three structures (any of which can be built using T terms according to the expansion rules discussed below). These structures represent the smallest possible set of words that can convey information, and are the building blocks for constructing more complex sentences. The simple structure helps to quickly translate into conversational natural language sentences; using the present invention, even compound sentences can be analyzed through module analysis of more basic sentence components (programs completed with better performance discussed below). Place into synonymous natural language. Basic structure 1 (BS1) is to place the descriptive word after the noun word to form the structure TD. BS1 sentences, such as "dog brown" and "Bill swim," can be easily translated into English sentences ‘’ the dog is brown ”(or the phrase“ the brown dog ") and “Bill swims”. 4UICKMAN / 200021TW; AND1P115.TW 1? 0 This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm)

Aw- Μ ------ (Please read the notes on the back before filling this page) Order --------- 548631 A7 B7 V. Description of the invention (\ 屮) ------- ----- · Equipment (please read the notes on the back before filling this page) BS2 is to add a connection word between two nouns to form the structure TCT. BS2 sentences like "dog eat food " can easily be translated into English sentences with the same meaning. BS3 is to add a reciprocal connective 'between two nouns to form a sequence represented by the structure TCT .... The sequence can be a single link Words, such as "Bob and Ted", can also be compound structures, such as "Bob and Tedand A1 and Jill" or " red or blue or green ". The following rules can be used to expand one or more of the basic structures mentioned above: Rule I: Add a descriptive word to the noun (T— > TD) #. Employee Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs print according to Rule I, Any language unit of the noun category can be expanded to the original item plus a new item of the descriptive category that modifies the original category. For example, "dog" becomes "dog big". Like all rules of the present invention, Rule I is not limited to a single noun (although this is the way BS1 sentences are formed); it can be applied to any noun, at the same time There is no restriction on the position of nouns in longer sentences. Therefore, according to rule I, TD1 — > (TD2) D1. For example, " dog big ”would become " (dog brown) big” (corresponding to an English sentence, " the brown dog is big ") 〇 When using consecutive adjectives, the order of adding new adjectives may or may not be important because all words modify T independently; for example, 4HICKMAN / 200021TW: AND1P115.TW 121 This paper size is in accordance with Chinese National Standard (CNS) A4 specification (210 χ 297 mm) 548631 Printed by A7 B7, Consumer Cooperative of Intellectual Property Bureau of the Ministry of Economic Affairs 5. Inventory (0V) " (dog big) brown ", The adjective " big " has distinguished this dog from other dogs, " brown " can describe characteristics that may not be known. When d-words are intransitive verbs, the order of addition is important. For example, in the TD sentence " dog run "(corresponding to " the dog runs" or "the running dog "), add the descriptive word" fast "to expand the sentence, and form" (dog fast) run "( Corresponds to " the fast dog runs "). If you want to express "the dog runs fast", you must use the descriptive word "run" to expand the TD sentence "dog fast" and become "(dog run) fast". 〇 Applying the expansion rule I to the structure BS2 will generate TCT— > ( TD) CT. For example, "dog eat food" becomes "(dog big) eat food". Rule I can also be applied to compound nouns in the form of TCT, so that the structure of BS3 becomes TCT -_ > (TCT) D. For example, "mother and father ', becomes "(mother and father) drive". In this way, multiple nouns can be connected or alternately combined for modification. Also note that the database contains transitive verbs, such as " drive ", as conjunctions and descriptive words. Another example is the verb " capsize", which can be considered as intransitive (" boat capsize ") and (&Quot; captain capsize boat ") verb. Rule Ila: Add conjunctions and other nouns (τ— > TCT) to nouns. According to rule Ila, any language unit in the noun category can be changed to two nouns before and after the connection, one of which is the original language unit. For example, " house "becomes" house on hill ". Apply the expansion rule Ila to 4HICKMAN / 200021TW; AND1P115.TW 122 This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 public love) ----- ------ · Equipment -------- Order --------- (Please read the precautions on the back before filling out this page) 548631 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. Invention description (丨 y) to BS1 will produce TD-> (TCT) D; for example, "gl〇〇my house " becomes " (house on hill) gloomy " ^ " the house on the hill is gloomy ”o Rule Ila can be used to join transitive verbs and their acceptors. For example, the compound word "mother and father" can be expanded to ‘’ (mother and father) drive car " ° rule lib: add a connection word and other nouns to a noun (T— > TCT). According to the rule lib, any language unit in the noun category can be changed to a conjunction followed by two nouns, one of which is the original language unit. For example, '' dog 'becomes' 'dog and cat'. Similarly, when the rule Rule Ila and the rule lib are applied, the noun can be a compound word composed of two or more nouns connected by a connection word. For example, the extended sentence " (john and bill) go-to market "satisfies rule Ila. Then applying rule I, this sentence can be further expanded to" ((j〇hn and bill) go-to market) together " ° Rule in: Add logical connection words and other description words (D-> DCD) to the description words. According to Rule III, the description word can be changed to a logical connection word followed by two description words, one of which is the original description word. For example, " big "becomes " big and brown". Applying Extended Rule III to BS1 will produce TD-> T (DCD); For example, " dog big "(equivalent to " the dog is big" or "the big dog") becomes 4HICKMAN / 200021TW; AND1P115.TW 123 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) (Please read the note on the back? Matters before filling out this page)

548631 Printed A7 B7 by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs V. Invention Description (\ called) "dog (big and brown)" (equivalent to 'the dog is big and brown' or 'the big brown dog' '') O A method of applying these rules to form an acceptable sentence according to the present invention is shown in Figure 38. FIG. 38 starts from a noun such as cat in step 3810, and forms any of three basic structures according to the expansion rules I, Ila, and lib, as shown in 3812, 3814, and 3816, respectively, to generate "cat striped" (BS1), ' 'cat on couch' (BS2) or 'cat and Sue' (BS3). Repeating the application of the extended rule Ila at 3818 and 3820 will generate TCI T1 — & (TC1 T1) C2 T2 or “(cat on couch) eat mouse) " and (TCI T1) C2 T2 — > ((TC1 T1) C2 T2) C3 T3 or "(((cat on couch) eat mouse) with tail)" and other formats. You can apply the expansion rule I to the T language unit at any point, as shown in 3822 (modify the original T, which is cat To produce " (happy cat) on couch ") and 3824 (modify " eat mouse "). You can also apply Rule III as shown in 3826 (further modify cat to produce " ((happy and striped) cat) on couch) ") and 3828 (further modification " eat mouse "). Extended rule I can be as 3812 Repeat the application as shown in 3, 3830 to further modify the original T (although as described in 3830, the descriptive word does not have to be an adjective). Using the extended rule Ila can show the action of the modified T (as shown in 3832), while rule I is available It is shown in the modification of newly added T (such as 3834). You can also use Rule I to modify (expand the present invention) the compound subject composed of rule nb, as shown in 3836. 4HICKMAN / 200021TW; AND1P115.TW 124 Paper size Applicable to China National Standard (CNS) A4 (210 X 297 mm) (Please read the precautions on the back before filling this page) · nnn 1 mmmte im a 0, i tm— ϋ · 1 1 · n · 548631 A7 B7 V. Description of the Invention (\ 25) The combination order of the language units has a great influence on meaning. For example, the extended sentence TC1 T1 "> (TC1 T1) C2 T2 can have many forms. The meaning of the sentence structure "cat hit (ball on couch)" is different from '' cat hit ball (on couch) '. The ball in the first sentence is of course on the sofa, and the ball in the back is the action that takes place on the sofa. The sentence "(john want car) .fast" indicates that the action should be completed quickly, and "(john want (car fast))" indicates that the car should drive fast. The following table is a more detailed example of the various expansion rules mentioned earlier , Illustrates the use of the present invention to express natural language: Table 8

Zairian health officials said 97 people have died from the Ebola virus so far. Jean Tamfun, a virologist, who helped identify the virus in 1976, criticized the government ’s quarantines and roadblocks as ineffective.

On

Saturday the quarantine on the Kikwith region was officially lifted. Health-official / s of zaire 4HICKMAN / 200021TW; AND1P115.TW 125 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) (please read the back first) Please fill in this page for the matters needing attention)-I I ---- I Order ---------, printed by the Employees 'Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 A7 B7 printed by the Employees' Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs Garment 5. Description of the invention (丨 u) * say * people 97 * dead * because-of * virus named ebola jean-tamfun be * virologist in zaire he help * scientist / s identify * virus named ebola * in 1976 jean-tamfun criticize * government of zaire he say * quarantine / s ineffective * and * roadblock / s ineffective government end * quarantine of * region named kikwit * on Saturday 4HICKMAN / 200021TW; AND1P115.TW 126 This paper applies the Chinese National Standard (CNS) A4 Specifications (210 X 297 mm) (Please read the precautions on the back before filling out this page) Binding ---- Order --------- 548631 A7 B7 V. Description of the Invention (丨 Clever) (Please first Read the back Note then fill Page) FIG. 39 is a specific embodiment of a representative hardware embodiment according to the present invention. As shown in the figure, the system includes the main two-way bus 3900. All system components communicate on the bus. The main instruction set for implementing the present invention and the database to be discussed later are stored in a large number of storage devices (such as hard disks or optical storage units) 3902. The system will also reside in the system main memory 3904 during system operation. These instructions and the present invention All functions are performed using the central processing unit (" CPU ") 3906. The user uses a keyboard 3910 and a position sensing device (such as a mouse) 3912 to connect to the system. The output of any of the above devices can be used to specify information or select a screen image 3914 The specific area above indicates the functions that the system should perform. The main memory 3904 printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs contains a set of modules responsible for controlling the operation of the CPU 3906 and interaction with other hardware components. Operating system 3920 is responsible for the management and execution of low-level system basic functions, such as memory configuration, file management, and operation of a large number of storage devices 3902. The analysis module 3925 implements a series of storage instructions at a higher level, instructing the execution of the main functions implemented by the present invention, Details are discussed below; the commands that define the user interface 3930 also It is possible to interact directly on the screen of the display 3914. The user interface 3930 generates text or graphic images on the display 3914, prompting the user to perform actions, and accepts user instructions from the keyboard 3910 and / or the position sensing device 3912. Main memory 3904 also contains definitions capable of storing the language units of the present invention. 4HICKMAN / 200021TW; AND1P115.TW 127 This paper size applies to the Chinese National Standard (CNS) A4 specification (210 X 297 mm) 548631 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs A7 B7 V. The partitions of many databases of the invention description (丨 z?) Are indicated by reference numbers 3935 ^ 3935 ^ 39353, 39354. These databases 3935 can be physically different (stored in the form of individual files on the storage device) Different memory partitions on 3902) or logically different (stored in a single memory partition in the form of a structured document, using addressing to identify many databases), each of which contains at least two languages Language units for specific categories. In other words, each database is organized into a table of The material column lists all language units of a specific category of a single language, so that each column can contain the same language units represented by different languages that the system can translate. In a specific embodiment, the nouns are placed in the database 39351, and A representative example of the single language (English) content of this database, that is, the content of a column in the multi-column work database is shown in Table 9. The connection word is placed in database 39352, and Table 10 is an example of its data column; Description Words are placed in database 39353, and table π is an example of its column; logical connectives (the simplest " and "and" or ") are placed in database 39354.

4HICKMAN / 200021TW; AND1P115.TW This paper size is applicable to China National Standard (CNS) A4 specification (210 X 297 meals) -----------, · install ------- order-- ------- (Please read the precautions on the back before filling this page) 548631 A / B7 V. Description of the invention (丨: 3〇) avalanche behavior brake ambassador capital Ministry of Economic Affairs Intellectual Property Bureau Employee Consumption Cooperative printed baby belgium brass captain amount back bell brazil car animal backpack belt bread ankle bag benefit breakfast cargo answer baker beverage breath carpenter ant balcony bicycle brick carpet apartment ball bill bridge carrot appetite banana billiard broom cash apple bandage bird brother cat appointment bank birth brush cattle barley birthday building cauliflc apricot barn bladder bulgaria cellar april barrel blanket bullet cemetery acchitect basket blood bus chain argentina

4HICKMAN / 200021TW; AND1P115.TW 130 ----------- install -------- order --------- ^ __ w— (Please read the note on the back first (Fill in this page again) This paper size is in accordance with Chinese National Standard (CNS) A4 (210 X 297 mm) 548631 A7 B7 V. Description of invention (\ w) Printed by the consumer property cooperative of the Intellectual Property Bureau of the Ministry of Economics, bath blouse butcher chair cheek copy dinner export germany cheese corkscrew direction eye gift chemistry corn disease face girl cherry cost dish factory glass chess cotton distance fall glasses chest couch document family glove chicken country dog farm glue child courage donkey father goat chile cousin door february god chin cow drawing ferry gold china cracker dream fig goose chocolate crane dress finger government Christmas cream driver fingernail grape church crib drum finland grapefruit cigar crime duck fire grass cigarette cuba dust fish greece circle cucumber eagle fist group

4HICKMAN / 200021TW: AND1P115.TW 131 (Please read the notes on the back before filling this page) This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) 548631 A7 -------- B7__Fifth, the invention description citizen cup ear flea guard clock curtain earring flood guest clothing czechoslov earthquake fly hair coat date egypt food hairdresser cockroach daughter elbow foot half cocoa day electricity football hammer coffee death elevator forest hand collar debt end fork handkerchief Colombia december enemy fox color decision energy france harbor comb degree engine friday harvest comfort denmark engineer friend hat competition 4HICKMAN / 200021TW AND1P115.TW 132 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) (Please read the precautions on the back before filling this page) Equipment ---- 111111. 548631 A7 B7 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs. 5. Description of invention (\ minutes) dentist england frog he computer departure entrance front head concert desert envelope fruit health condition dessert ethiopia funeral heart connection diarrhea europe game heel conversation dictionary excuse garden here digestion exhibition cook dining- exit copper room holland key honey kidney horse kind horse-race king hospital kitchen expense luggage lunch lung highway hole gauge holiday movie pain mushroom painting mustard pair garlic gasoline machine nail pakistan magazine nail-file

4HICKMAN / 200021TW; AND1P115.TW 133 (Please read the phonetic notes on the back before filling out this page) Loading ---- Order --------- · This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) A7 548631 B7 V. Description of the invention (outside) Printed by pancake hotel knee magic name panic hour knife maid nature pants house kuwait mail neck paper hungary lace malaysia necklace parachute husband ladder malta needle parents I lake man neighbor parking ice lamb map nepal part ice-cream language march netherlands partridge iceland lawyer market new- passport idea lead marriage Zealand pea import leaf match newspaper peace india leather mattress nicaragua pear Indonesia lebanon may nigeria peasant information leg meat night pen ink lemon medicine noodle pencil 4HICKMAN / 200021TW; AND1P115.TW 134 ----------- install -------- order -------- (Please read the Note the heart note, then fill out this page) This paper size applies Chinese national standards CNS) A4 specification (210 X 297 mm) 548631 A7 B7 Printed by the Consumer Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs. 5. Description of invention (\ π) insect letter insurance liberia interpreter library invention libya meeting noon people iran iraq license life ireland light melon north- pepper member america persia memorial north-pole peru metal norway pharmacy mexico nose Philippines middle november physician iron light-bulb milk number piano island lightning minute nurse picture israel mistake mistake nut Pig it linen monday oak pigeon italy lion money oar pillow january lip monkey oats pilot japan liquid month October pin jewel liver moon office pine-tre (job living-: room morning oil pipe joke lobster morocco olive plant

4HICKMAN / 200021TW; AND1P115.TW 135 (Please read the note on the back before filling out this page) • The size of the paper is applicable to China National Standard (CNS) A4 (210 X 297 mm) A7 548631 B7 V. Invention Description (丨%) Printed by the Consumer Cooperative of Intellectual Property Bureau of the Ministry of Economic Affairs, jordan lock mosquito onion platform juice look mother orange play July loom mountain ore playing- june love mouse ox card kenya luck mouth package pleasure plum room skin story tin pocket root skis stove tire poison rope sky street toast poland rubber sled student tobacco police- rumania smell subway today officer russia smoke sugar toe porter rust snake summer toilet portual saddle snow sun tomato post-office saddness soap Sunday tomorrow postcard safety socks surprise tongue pot saftey-belt soda swamp tool potato sailor soldier Sweden tooth powder salt solution Switzerland toothbrush prison sand son syria top 4HICKMAN / 200021TW; AND1P115.TW 136 ------------ # 装 ---- (Please read first Please fill in this page again before ordering) Order --------. Line · Γ This paper size is applicable to Chinese National Standard (CNS) A4 (210 X 297 mm) 548631 A7 B7 i. Description of the invention (丨V]) problem Saturday property sauce purse saudi-quarter arabia queen squsage question scale rabbit scarf song sound table towel tail tailor town toy soup south-africa taste train south- tax tree trip america tea south-pole teacher trouble radio school soviet- telephone truth television tuesday tent rag science union (Please read the notes on the back before filling in this page) rain raincoat rat scissors space Scotland spain screw spice tunisia -------- Order --------- Ministry of Economic Affairs Intellectual Property Bureau Employee Consumer Cooperatives Print Yiyi spoon test turkey thailand tv-show theater typewriter razor sea receipt self spring they umbrella record- September staircase thief uncle player shape stamp thigh united-

4HICKMAN / 200021TW; AND1P115.TW 137 This paper size applies to Chinese National Standard (CNS) A4 (210 X 297 mm) 548631 A7 B7 V. Description of invention (other) refrigerator Intellectual Property Bureau, Ministry of Economic Affairs, Employee Consumption Cooperative, printed by she star thing states religion sheep starch thirst Uruguay rent shirt station thread us restaurant shoe steak throat vaccination result shoulder steel thumb vegetable rice side stick thunder velvet ring signature stock- thursday Venezuela risk silk market ticket victim river silver stomach tie view rocket sister stone tiger village roll situation store time vinegar roof size storm timetable violin voice water weight window work waiter we wheat winter year wall weather where? woman yesterday war wedding who? wood you waste Wednesday wife wool Yugoslavia watch week wind word 4HICKMAN / 200021TW: AND1P115.TW 138 --- ---------------- Order --------- line (please read the precautions on the back before filling this page) This paper size applies Chinese national standard (C NS) A4 specification (210 X 297 mm) 548631 A7 B7 Five inventions printed by the Intellectual Property Bureau of the Ministry of Economic Affairs, printed by the consumer co-operatives of the invention 10 Formable-to call from mix shoot about called from more-than should above capsize fry move sing across capture give near smell afraid-of carry go-in need speak after catch go-through occupy steal against cause go-to of sting allow change hang on stop answer climb hate outside study arrest close have pay take arrive-at cook hear play teach ask count help prepare ： throw at cut hit print to bake deal-with hunt promise touch be decrease if prove translate because defeat in pull try 4HICKMAN / 200021TW; AND1P115.TW 139 This paper standard is applicable to China National Standard (CNS) A4 specification (210 X 297 mm) -------------------- Order --------- line (Please read the precautions on the back before filling this page) 548631 A7 B7

V. Description of the invention (丨 VO becomes deliver in-front-of push turn-off before discuss in-order-to put turn-on begin down include read under behind drink increase reduce understand believe drive kill refuse until bet drop kiss remember use betray eat know repeat value between examine learn ride visit blame explain leave roast want bother find like say wash break finish live-in see while bring fix look-for sell win burn for made-of send with but for make sew work-for buy forget meet shave write -------------------- Order --------- line (Please read the precautions on the back before filling in this page) Intellectual Property of the Ministry of Economic Affairs 4HICKMAN / 200021TW; AND1P115.TW 140 printed by the Bureau ’s Consumer Cooperatives. This paper size applies to Chinese National Standard (CNS) A4 (210 X 297 mm) 548631 Α7 Β7 V. Description of the invention (\ + \) Table 11 Descriptive Word Economy Printed by the Consumer Cooperatives of the Ministry of Intellectual Property Bureau, broad clean flat long round absent clear fly malignant run again cold forbidden maybe sad agree complain for eign mean safe alive continue fragile more short all correct free much sick almost cough fresh mute similar alone crazy fun mutual sit also cry funny my sleep always curious glad nervous slow angry damp good neutral slowly another dangerous goodbye never small any dark green new smile argue dead grey next soft artificial deaf grow nice some automatic decrease guilty north sometimes 4HICKMAN / 200021TW; AND1P115.TW 141 (Please read the precautions on the back before filling in this page) -If order --------- line · this paper Standards are applicable to China National Standard (CNS) A4 specifications (210 X 297 mm) 548631 A7 B7 V. Description of the invention (\ o available deep hang backward defective happen bad different happy bashful difficult hard beautiful not sour now often south special okay stand Ministry of Economic Affairs Intellectual Property Bureau employee consumer cooperative printed clothes dirty healthy old strong begin drop heavy open sweet black drown hungry our swim blind dry illegal permitted blond early important pink tall blue east increas e play thanks boil easy intelligent please there boring empty interesting poor thick born enough jealous portable thin brave expensive kiss possible thinl broken expire large previous tired

4HICKMAN / 200021TW; AND1P115.TW 142 (Please read the precautions on the back before filling this page) This paper size is applicable to China National Standard (CNS) A4 (210 X 297 mm) A7 548631 B7 V. Invention description brown extreme last quiet together burn far late red too-much capsize fast laugh rest transparent careful fat lazy rich travel change few left right ugly cheap first legal ripe upstairs urgent warm wet worry young wait weak white wrong your walk west why? yellow (please read the first (Please note the phonetic matters and fill in this page again.) 0-丨 Line · The employee's consumer clothing cooperative printed clothing input buffer 3940 of the Ministry of Economic Affairs receives the user's input sentence through the keyboard 3910. It is preferably composed according to the structure of the present invention and the format is The following description. At this time, the analysis module 3925 starts to check whether the input sentence conforms to the structure. Module 3925 will process the single-language unit of the input sentence interactively, and search the database to find the item corresponding to each language unit in the specified language and the corresponding item in the target language. The analysis module 3925 translates the sentence, replaces the input item with the target language item, sends the translation result to the output buffer 3945, and displays it on the screen of the display 3914. Although the module of main memory 3904 is discussed separately, it is only for the sake of clear expression; as long as the system can perform all necessary functions, the module is in Department 4HICKMAN / 200021TW; AND1P115.TW 14 ^ This paper standard applies Chinese national standards ( CNS) A4 specification (210 X 297 mm) A7 548631 ____B7 ____ 5. Description of the invention (丨 蚪) The distribution method in the system and its programming structure are not important. In order to make the module 3925 easy to analyze, the input sentence is preferably formed in a unique format that can be easily processed, so as to facilitate the direct identification of individual language units and simple verification to determine that a series of units are reasonable sentences that comply with the expansion rules of the present invention. In one application (straight), each language unit of a sentence is displayed on a separate line. If an expansion rule is applied, the asterisk (*) is used to mark the expanded part; that is, the basic sentence structure is connected together with * to form a longer sentence. For example, in the drawing of each item in Figure 1, cat striped * hit * ball red represents the results of steps 132 and 134. Or you can use the algebraic (horizontal) format to represent the sentence, and use the brackets to expand the words to identify the extended parts: (cat striped) hit (ball red) 4HICKMAN / 200021TW; AND1P115.TW 144 This paper standard applies to Chinese National Standard (CNS) A4 specification (210 X 297 mm) ------------ dr (Please read the notes on the back before filling this page) Order --------- Intellectual Property of the Ministry of Economics Printed by the Bureau ’s Consumer Cooperatives A7 548631 B7__ 5. Description of the Invention (H5) In any case, the user ’s input will be treated as a character string, and using standard string analysis routines, module 3925 will identify each Language units and extension points. These units are then compared with a template corresponding to the allowed expansion rules, the sentences are verified, and then a database search and translation is performed. If the sentence does not comply with the rules of the present invention, the module 3925 alerts the user via the display 3914. According to the above various forms of expression, English singular nouns will have n / s " at the end to form plurals (such as " nation / s "). For other languages, plurals will be formed using the most common method; for example, French and English Add W as well, but in Italian it will add " / i ". Numbers are represented by numbers. Alternatively, the analysis module 3925 can be configured to process unformatted input sentences. When module 3925 is executed, it will search each input word (or a group of words) in database 3935, and then construct a sentence representation according to the language category of the words, that is, replace each unit with a language category symbol. Module 3925 then evaluates whether the generated category order is generated in accordance with the allowed expansion rules. If so, the language units are classified for search and translation. The output is provided in an unstructured format corresponding to the input, or in one of the previously mentioned formats. The latter is the preferred output form, because if only replaced, strings in one language can hardly correspond to strings in another language; usually the language units are isolated, and the extended output format will be easier to understand. 4HICKMAN / 200021TW: AND1P115.TW 145 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) -------------------- Order- ------- " mp (Please read the notes on the back before filling this page) Printed by the Consumer Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 a? ____ Β7 Description of the invention (丨 from) This invention can be added to other Features to simplify the job. For example, the aforementioned ‘uses a period at the end of a word to distinguish words with multiple meanings; the number of periods after a particular word meaning represents an arbitrary choice. Therefore, another database 3935 can constitute a dictionary with multiple meaning words, and after each definition, set each word meaning format that the present invention can recognize. The user interface 3930 interprets the user's key press on the definition as selecting the definition, and then inputs the appropriate encoding of the word into the input buffer 3940. Similarly, because the consideration of the scale and speed of the operation economy limits the size of the overall database, one of the databases 3935 can be set as a synonym dictionary to provide an unrecognizable input word that is closest to the language unit that the present invention can recognize. If the operation attempt of the analysis module 3925 fails, and you want to find the words in the database, you can schedule the module 3925 to look up the synonym database 3935, and then return a list of words in the language database. Module 3925 may also contain specific utilities that can identify and correct (for example, after user approval) common errors in sentence structure. For example, the present invention usually uses the verb " to have1 'to indicate the possessive of the specified person; for example, `` PauPs computer is fast " This sentence will be expressed (in algebraic format) as `` paul have (computer fast)' 'or " (computer of paui) fast "; if unnamed, generic possessive pronouns are used (eg " (computer my) fast "). Therefore, the module 3925 can be configured to recognize a structure such as' 'Paul's', and return an appropriate structure according to the present invention.

4HICKMAN / 200021TW; AND1P115.TW This paper size is applicable to Chinese National Standard (CNS) A4 (210 X 297 mm) -------------— (Please read the note on the back of the phonetic note first, then (Fill in this page) Order --------- line _ · Printed by the Employees ’Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs 548631 Α7 Β7 Printed by the Employees’ Cooperatives of the Intellectual Property Bureau of the Ministry of Economic Affairs It can be seen that the above represents a convenient and fast method of translating between multiple languages. The words and expressions used here are for illustration only, and the application of the present invention is not limited to this, and these words and expressions are used The formula does not preclude the display and discussion of any equivalent application of features or parts. 'We determine that there are many possible variations of applications within the scope of this application. For example, in a general-purpose computer, using appropriate software instructions or hardware circuits' or The combination of software and hardware can implement the various modules of the present invention. Various specific applications have been discussed previously, but it should be understood that these are illustrative examples and are not limitations of the application of the present invention Therefore, the application scope and scope of the preferred embodiments are not limited by any of the foregoing exemplary embodiments, but should be defined according to the following patent application items and their equivalent applications. 0 4HICKMAN / 200021TW; AND1P115.TW 147 sheets Standards are applicable to China National Standard (CNS) A4 specifications (210 X 297 mm) -------------------- Order --------- line (please (Read the notes on the back before filling out this page)

Claims (1)

548631 Α8 Β8 C8 D8 Member of the Intellectual Property Bureau of the Ministry of Economic Affairs, printed by Consumer Cooperatives 6. Revision of the scope of patent application 2002/06/24 1. A method for speech recognition on a network to access data, including the following steps: ( a) prompt the user to provide a voice sample through a network; (b) receive the voice sample from the user through the network; (c) retrieve the user's registration information through the network, including a voice scan ; (D) comparing the voice sample with the voice scan to verify the user's identity; (e) if the user's identity is verified, granting the user's data access rights over the network; and (f ) If the user ’s identity is not verified, the user is denied access to the data. 2. The method according to item 1 of the scope of patent application, further comprising the step of recording a user's voice to establish the voice scan and store the voice scan. 3. The method according to item 1 of the patent application scope, wherein the voice scan includes more than one word spoken by the user. 4. The method described in item 3 of the scope of patent application, wherein if a first word cannot verify the identity of the user, the user is prompted to say another word. 4HICKMAN / 200021TW; AND1P115.TW This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) 151 > -------------------- Order --------- line (please read the notes on the back before filling this page) 548631 Printed by the Consumer Cooperatives of Intellectual Property Bureau of the Ministry of Economic Affairs Λΰ Β8 C8 D8 2/24 Case No .: 89117686 5. According to the method described in the first patent application scope, if the identity of the user is verified, the user is granted access to a web site on the Internet. 6. The method according to item 1 of the scope of patent application, further comprising the step of recording the user's voice sample and the recording time and date of the voice sample. 7. —A computer-readable medium with a program code executable by a computer, a method for performing speech recognition on a network to access data, the method comprising: (a) retrieving a User registration information, including a voice scan; (b) · Prompting the user to provide a voice sample on the network; (c) Receiving the voice sample from the user on the network; | (d) Comparison The voice sample and the voice scan to verify the user's identity; (e) if the authentication passes the user's identity, grant the user access to data on the network; and (0 if the authentication fails the user's Identity, the user is denied access to the data. 4HICKMAN / 200021TW; AND1P115.TW This paper size applies the Chinese National Standard (CNS) A4 specification (210 X 297 public love) 548631-'. M ΛΟ Β8 C8 D8 L One….… .One ............ 6 / Application for Patent Scope Replacement Page 20003/2/24 Case No. 89117686 8. If the scope of patent application is The computer program according to item 7, further comprising a piece of code that records a voice of the user to create the voice scan and stores the voice scan. 9. The computer program according to item 7 of the scope of patent application, wherein the voice scan includes more than one word spoken by the user. 10. According to the computer program described in item 9 of the scope of patent application, if the identity of the user fails to pass the verification of a first word, the user is prompted to say another word. 11. The computer program described in item 7 of the scope of patent application, if the identity of the user is verified, grants the right to access a website on the Internet. .. 12. The computer program as described in item 7 of the scope of patent application, further comprising a piece of code for recording the voice sample of the user and the recording time and date of the voice sample. 13. A system for speech recognition to access data on a network, including: 4HICKMAN / 200021TW; AND1P115.TW 153 This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) --- -------------- (Please read the notes on the back first). • Cast. Intellectual Property Bureau Employee Consumer Cooperatives, Ministry of Economic Affairs, PR 54863 Old%, Intellectual Property Bureau Employee Consumer Cooperatives, Ministry of Economic Affairs Printed 6. Amendment to the scope of patent application 2002/06/24 (a) Logic to retrieve a user's registration information on a network, the registration information includes a voice scan; (b) Prompt the user on the network Logic to provide a voice sample; (c) logic to receive the voice sample from the user on the network; (d) logic to compare the voice sample with the voice to verify the user's identity; (e) if The logic of verifying the identity of a user and granting the user access to data on the network; and the logic of denying the user access to data if the verification fails. 14. The system according to item 13 of the scope of patent application, further comprising logic for recording a voice of the user to establish the voice scan and storing the voice scan. 15. The system according to item 13 of the patent application scope, wherein the voice scan contains more than one word spoken by the user. 16. The system according to item 15 of the scope of patent application, wherein if the identity of the user fails the verification of a first word, the user is prompted to say another word. 17. · The system described in item 13 of the scope of patent application, if the verification passes 4HICKMAN / 200021TW; AND1P115.TW 154 --------------------- Order --- ------ Line (Please read the notes on the back before filling this page) This paper size applies to China National Standard (CNS) A4 (210 X 297 mm) 548631 Travel I A8 B8 C8 D8 VI. Patent Application Scope Amendment 2002/06/24 The identity of the user grants access to a website on the Internet. 18. The system according to item 13 of the scope of patent application, further comprising logic for recording the voice sample of the user and the recording time and date of the voice sample. (Please read the notes on the back before filling in this page) f—— Order --------- Printed by the Consumers Cooperative of the Intellectual Property Bureau of the Ministry of Economic Affairs 4HICKMAN / 200021TW; AND1P115.TW 155 This paper size applies China National Standard (CNS) A4 Specification (210 X 297 mm) 548631 U—0CSIZCO Λ * οBu House) /