JP2005181358A - Speech recognition and synthesis system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech recognition and synthesis system that enables equipment control by speech input close to a conversation between persons by making it possible to input a voice command consisting of a document that people daily use. <P>SOLUTION: In addition to functions of a conventional speech recognition/synthesis device, this system is provided with a history database 107 which stores histories of operations of prescribed equipment by users, a means 108 of analyzing the database to generate a script, a means of specifying a user based upon the recognition result of a speech recognition processing means 102 when there are a plurality of users, and means (101, 110, 107, and 108) of analyzing the operation histories stored in the history database, user by user, to relate specified keywords to each other and generating or updating a script of an operation procedure of the prescribed equipment from the related information. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a system that uses voice recognition or voice synthesis to control a device by allowing a user to input a command by voice in an interactive manner.

  Currently, in the field of computers and electronic devices, there are devices that can recognize a user's voice command and control the device, and can perform character input by voice. In such devices, not only the voice is recognized, but also the voice is synthesized and the confirmation of the command content and the start / end of the operation are transmitted to the user by voice.

A series of device control operations will be described by taking car navigation (hereinafter referred to as car navigation) as an example.
An example of the exchange between the user and the car navigation when inputting the destination to the car navigation is shown below.
(Car navigation system) "Please enter your destination"
(User) “Tokyo”
(Car navigation) "Which is Tokyo?"
(User) “Setagaya Ward”
(Car navigation system) "Which is Setagaya Ward?"
...
By exchanging information with the device, the user can input the destination by voice without using a hand or the like while driving, and can cause the car navigation to display a map of the destination or search for a route to the destination. it can. In recent years, with the development of speech recognition technology, it has become possible not only to recognize words but also to recognize consecutive words. For example, in the destination entry, the text "Go to Setagaya 1-chome, Setagaya-ku, Tokyo" is entered. By doing so, the route search to the destination can be performed.

By the way, in such a system capable of operating a device by voice input, the definition of the exchange between the user and the device, that is, what kind of word is input and what device operation is performed next, for example, A scenario of an operation procedure such as waiting for input of a word, starting execution of a route search, and simultaneously responding to the user by speech synthesis becomes important. There is also a method of implementing this scenario in an application programmatically, but currently, a text file-based script is defined, and the script file is read and executed sequentially by a browser program that has a speech recognition engine or speech synthesis engine. There is technology. That is, in the technology disclosed in the following Patent Document 1 (title of invention: voice command system, voice command device, voice command method, and voice command program), this script is described on an HTML basis. In the technology disclosed in Non-Patent Document 1 below, the W3C standardizes voiceXML that describes this script on an XML basis.
Also, a modified speech recognition / synthesis system that handles a specific natural language is widely known as described in Patent Document 2 below.
JP 2002-366344 A Japanese translation of PCT publication No. 2003-521750 (claims 77 and 78) Voice eXtensible Markup Language (VoiceXML) version 1.0 W3C Note 05-May-2000 (http://www.w3.org/TR/2000/NOTE-voicexml-20000505)

In the speech recognition / synthesis system using the script file described above, device operation is performed in accordance with procedures defined in script files that are prepared for various user operation scenes prepared in advance. Therefore, when performing the same operation, it is necessary to exchange a series of sounds. For example, if you want to go again to “Setagaya-ku, Setagaya-ku, Tokyo” as in the conventional example, you have to enter “Setagaya-ku, Setagaya-ku, Tokyo” again. Needless to say, it is not convenient for the user to make the user repeat the same thing over and over. In addition, when the device asks the same thing many times in this way, it becomes far from the natural conversation between humans, and it has also hindered the spread of speech recognition and synthesis systems.
Further, the system described in Patent Document 2 is a speech system and cannot be applied to command transmission.

  Therefore, in the present invention, processes corresponding to words such as “usual” and “previous” are frequently used in a script file as humans often use on a daily basis based on a user's past voice input history and device operation history. By reflecting it, it is possible to input voice commands by sentences that humans use on a daily basis, such as “go to the usual place” or “go to the previous shop”, and device control is possible with voice input that is closer to the conversation between humans An object of the present invention is to provide a speech recognition and synthesis system.

The present invention includes a speech recognition processing unit and a speech synthesis processing unit as means for achieving the above object, and the contents of a script in which an operation procedure in a controlled device corresponding to a voice command input from a user is defined. In the speech recognition and synthesis system for sending an operation command to be executed to the controlled device and causing the controlled device to execute a desired operation by the user's voice command,
Means for identifying the user based on a recognition result in the voice recognition processing unit when there are a plurality of the users;
A database for storing the operation history of the controlled device by a plurality of users in association with each user specified by the means for specifying;
Means for analyzing the database and generating a script;
Means for analyzing the operation history stored in the database for each specified user, associating specific keywords with each other, and generating or updating a script of an operation procedure of the controlled device from the associated information And
The present invention provides a speech recognition and synthesis system characterized by having the above.

  In the speech recognition / synthesis system of the present invention, a user's past operation history is analyzed, and a keyword related to the operation with the highest number of times, a keyword related to the last operation performed, and “special” or “previous” special keywords. By associating with a keyword, it becomes possible to respond to the command input from the user such as “ordinary XXX” and “previous XXX”. You do n’t have to.

In addition, by extracting speaker characteristics during voice recognition, identifying the speaker, and generating a script that reflects the history of each user, commands such as “usual” and “previous” for each user Can be optimally operated.
In addition, the ability to use command input that is similar to text used by humans on a daily basis can provide the effect of reducing the sense of resistance to talking with the machine.

The speech recognition / synthesis system of the present invention analyzes the past operation history for each user, and the keyword related to the most frequently used operation, the keyword related to the last operation, and special such as “ordinary” and “previous”. By associating with various keywords, it is possible to respond to command inputs such as “ordinary XXX” and “previous XXX” from the user.
FIG. 1 is a block diagram showing an embodiment of a speech recognition / synthesis system according to the present invention. In the present embodiment, car navigation is assumed as an example of the operation target device.

  The voiceXML interpreter 101 reads and executes the voiceXML script 109. The voiceXML script 109 is stored in a rewritable medium such as a hard disk or RAM in the system. The voice recognition engine 102 recognizes the voice input by the user using the microphone 104 and transmits it to the voiceXML interpreter 101 as a character string. In addition, the voice recognition engine 102 identifies a speaker and transmits the speaker information to the user management unit 110. The speech synthesis engine 103 converts the character string transmitted from the voiceXML interpreter 101 into speech, outputs it to the speaker 105, and responds to the user.

  For example, if the voiceXML script 109 outputs a voice saying “Where is the destination?”, Waits for voice input from the user, and after inputting the destination, it is described that the route search is performed for the destination. The interpreter 101 first transmits a character string “Where is the destination?” To the speech synthesis engine 103 to output the speech.

  Thereafter, the process waits for a character string from the speech recognition engine 102. When a character string is transmitted from the speech recognition engine 102, the word is interpreted and a route search instruction is issued to the device operation processing unit 106. The device operation processing unit 106 stores the operation history and information from the user management unit 110 in the history database 107 after completing processing such as displaying the route search result to the user. Thereafter, the history analyzer 108 starts analysis based on the information stored in the updated history database 107. For example, the user searches for the place where the route was searched last, the place where the number of past searches is the largest, and the like. A similar search is performed for each attribute such as genre. Based on the result, the voice XML script 109 is updated by associating with a special keyword such as “ordinary store”, “store visited last time”, and “usual place” for each user. After that, when the voiceXML script 109 is executed again, the voiceXML interpreter 101 receives the speaker information and the voiceXML script 109 from the user management unit 110 when a command such as “ordinary store” or “previous store” is input from the user. The route search process corresponding to the user is performed with reference to the contents of the above.

  The history database will be described in detail using an example of the history database in FIG. 2. The history database includes a search location, a search date, attribute 1, attribute 2, search frequency (search history), user name (user information), and the like. The item is saved. Referring to this database, as a keyword associated with “ordinary store”, in the case of user A, “XX convenience” having the largest number of searches among search locations of items having the attribute “store” is applicable. . The keyword associated with “the previous store” is “XX restaurant” with the latest search date and time among search locations having the attribute “store”. On the other hand, in the case of the user B, “BB super” having the largest number of searches corresponds to the search place of the item having the attribute “store”. The keyword associated with “the previous store” is “ramen BB” having the latest search date and time among search locations having the attribute “store”.

  In this way, the history database is analyzed, and further, from the association, when a voice command “ordinary store” is issued by the user, “XX convenience” for user A, and “BB supermarket” for user B. Update the existing voiceXML script to the voiceXML script defined to guide you.

FIG. 3 is an example of a voiceXML script before update. After outputting a voice saying "Where is your destination?", It waits for voice input from the user, and the route search process after input starts.
On the other hand, FIG. 4 shows an example of the updated voiceXML script. If a keyword such as “ordinary store” or “previous store” is input after the user inputs voice, the operation history for each user is displayed. It is defined to behave accordingly.
By reflecting the past operation history in the voiceXML script in this way, after the device asks "Where is the destination?", The user responds "Normal store" or "Previous store" It is possible to guide the store according to the past operation history.

  In the present embodiment, a route search for car navigation is assumed, but the present invention can be applied to other commands, and can also be applied to systems using speech recognition synthesis in other fields. In this embodiment, voiceXML is used as a script. However, HTML and other scripts can also be used.

  In the embodiment shown in FIG. 1, a device operation processing unit 106 is provided, and a part of a device whose operation is controlled using the speech recognition / synthesis system of the present invention is incorporated. Yes. However, in the case of a configuration that does not include a part of the device to be controlled, a control signal for controlling the device to be controlled is sent instead of the device operation processing unit 106 in FIG. An interface or the like for receiving a signal indicating the state can be provided.

  A portion excluding the microphone 104 and the speaker 105 in the embodiment of the present invention shown in FIG. 1 is a CPU (Central Processing Unit), a RAM, a ROM, an interface, a bus line, etc. that are components of the computer. Can be configured. Therefore, each function in the apparatus according to the present invention described in the above embodiment can be embodied as a computer program and executed by a computer. Such a computer program can be supplied by being recorded on a predetermined recording medium and taken into the computer, or can be sent via a communication network such as the Internet and taken into the computer.

  The speech recognition / synthesis system according to the present invention can respond to command input from the user such as “ordinary XXX” and “previous XXX”, and repeatedly input the same voice command to the user, such as inputting the same address many times. Since it is not required, it is useful not only in the car navigation system described in this embodiment but also in various systems in which a user operates a device using speech recognition synthesis.

It is a block diagram which shows embodiment of the speech recognition synthesis system of this invention. It is a figure which shows an example of the structure of the history database used for embodiment of the speech recognition synthesis system of this invention. It is an example of the voiceXML script used for the embodiment of the speech recognition and synthesis system of the present invention (before update). It is an example of the voiceXML script used for embodiment of the speech recognition synthesis system of this invention (after update).

Explanation of symbols

101 voiceXML interpreter (configures a means for generating or updating a script together with a user management unit, a history database, and a history analyzer)
102 Voice recognition engine (voice recognition processing unit)
103 speech synthesis engine 104 microphone 105 speaker 106 device operation processing unit 107 history database 108 history analyzer (means for generating a script)
109 voiceXML Script 110 User Management Department

Claims (1)

The controlled device includes a voice recognition processing unit and a voice synthesis processing unit, and causes the controlled device to execute an operation command for executing the contents of a script in which an operation procedure in the controlled device corresponding to a voice command input from a user is defined. In a speech recognition and synthesis system for sending and executing a desired operation on the controlled device according to the user's voice command,
Means for identifying the user based on a recognition result in the voice recognition processing unit when there are a plurality of the users;
A database for storing the operation history of the controlled device by a plurality of users in association with each user specified by the means for specifying;
Means for analyzing the database and generating a script;
Means for analyzing the operation history stored in the database for each specified user, associating specific keywords with each other, and generating or updating a script of an operation procedure of the controlled device from the associated information And
A speech recognition and synthesis system comprising:

Images