KR100309219B1 - Network-based speaker learning and verification method and system thereof - Google Patents

Abstract

The present invention relates to a network-based speaker learning and speaker identification method and apparatus, and more particularly, to a letter-directed speaker learning and speaker identification using a group of words that well represent the individual characteristics of the speaker.

Instead of using a separate password to log in to a site providing information using the Internet, the voice characteristics of each individual are used among the unique characteristics of each individual.

In addition, in order to increase the speaker recognition rate and impersonate defense due to recording, a word group composed of words well representing the individual characteristics of the speaker according to the gender and age of the speaker is used.

Since the number of simultaneous users on the Internet is unpredictable, it is designed to prevent the load from being concentrated in one place even when many users access the system at the same time.

A plurality of words may be learned, and each word may be selected and used to identify the speaker by selecting one word from the learned words in consideration of the time, the speaker's state, and the connection location.

Description

Network based speaker learning and speaker identification method and device {NETWORK-BASED SPEAKER LEARNING AND VERIFICATION METHOD AND SYSTEM THEREOF}

The present invention relates to a network-based speaker learning and speaker identification method and apparatus.

With the development of computers, society has become increasingly complex and requires the rapid exchange of information. In particular, with the advent of the Internet, all information and communication technologies must coexist with the Internet to become a society that is recognized for its value. Recently, sites that provide information using the Internet (Site) have been increasing in the number of sites that provide information only to members by operating the site in order to provide high-quality information to a specific person. In the case of a site operated by a membership system, the site can be used after inputting a membership number and password. However, these login methods have the following problems.

First, there is a problem that the password is easily forgotten or known to others by hacking or the like, and thus is vulnerable to theft by others.

Second, in recent years, since an individual has 3 to 5 member numbers and passwords, it is often impossible to log in to the site due to the forgetting or confusion of passwords. There is a problem that inhibits the activation of the use of the site. There is a method of authenticating the user by using the biological characteristics of the user, for example, a user authentication method by speaker recognition, iris recognition, fingerprint recognition, and the like. In particular, recently, a user authentication method using a speaker recognition method has appeared. However, the conventional user authentication method using the speaker recognition method recognizes the user by causing the user to pronounce only one word. Since only one word is pronounced, there is a problem that another person can easily steal the pronunciation element of the single word. Further, according to the conventional user authentication method by pronunciation of a single word, it is appropriate to the user's state. There is a problem that cannot be coped. Specifically, when a user pronounces a single word (for example, 'my house') when the user is healthy, the pronunciation elements of the single word are learned and stored in advance. After that, if the user is caught by the throat and pronounces the word ('my house'), the user may not be authenticated despite being a real user. Generally, according to the user authentication method using a speaker recognition method, the speaker learning process is performed. For example, the process of extracting a speaker and / or extracting a speaker component requires a large amount of processing. According to the user authentication method using the conventional speaker recognition method, each process requiring the large amount of processing in the server is dedicated. As a result, there is a problem that the load is concentrated on the server and the user authentication process cannot be performed smoothly due to server down. In addition, in a conventional electronic commerce, after a customer makes a certain order and a contract is established, There was a lack of countermeasures if the customer had not known or denied the establishment or ordering of the contract.

Accordingly, an object of the present invention has been made in view of the above problems.

A specific object of the present invention is a user authentication method including speaker learning and speaker identification using the voice characteristics of each individual without the need for installing additional equipment other than the basic computer environment among the unique characteristics that each individual has for user verification. And to provide an apparatus.

Another object of the present invention is to provide a user learning method and apparatus including a speaker learning and a speaker using a group of words composed of a plurality of words for improving the speaker recognition rate and impersonating the defense due to recording. User authentication method including a letter-directed speaker learning and speaker identification using a word group consisting of words well representing the individual characteristics of the speaker according to the sex, age, etc. of the speaker to improve the speaker recognition rate and impersonation protection due to recording; and To provide a device.

It is still another object of the present invention to provide a user learning method including distributed speaker design and speaker, which prevents the load from being concentrated in one place even when many users are connected to the system because the number of simultaneous users is unpredictable on the Internet. It is another object of the present invention to provide a device. In a case where a contract is established by a customer in an e-commerce transaction, a voice signal corresponding to the order is received, processed, and stored, thereby allowing the customer to contract. In the case of noting or denying the fact of establishment or ordering, it is possible to enhance the stability of electronic commerce by restoring the voice signal to the customer.

1 is a schematic configuration diagram of an apparatus to which the present invention can be applied.

Figure 2a is another schematic configuration diagram of a device to which the present invention can be applied.

Figure 2b is another schematic configuration diagram of a device to which the present invention can be applied.

3 is a flow chart showing a speaker learning method according to an embodiment of the present invention.

4 is a schematic flowchart illustrating a preprocessing process according to an exemplary embodiment of the present invention.

5 illustrates a distributed processing scheme, in accordance with a preferred embodiment of the present invention.

6 is a flow chart schematically showing the learning process according to an embodiment of the present invention.

7 is a flowchart showing a speaker learning method according to another embodiment of the present invention;

8 is a flow chart showing a speaker identification method according to an embodiment of the present invention.

9 is a schematic flowchart showing a speaker identification process according to an exemplary embodiment of the present invention.

10 is a flow chart showing a speaker identification method according to another embodiment of the present invention.

11 is a flowchart showing a speaker learning method according to another preferred embodiment of the present invention.

12 is a flowchart showing a speaker learning method according to another preferred embodiment of the present invention.

13 is a flow chart showing a speaker identification method according to another embodiment of the present invention.

14 is a flow chart showing a speaker identification method according to another embodiment of the present invention.

<Explanation of symbols for the main parts of the drawings>

101... Server 103... Client

105... Web site database 107... Voice signal database

201... Network server 203... Speaker recognition server

In order to achieve the above object, according to an aspect of the present invention, in the network-based user authentication for authenticating a user connected by using a microphone in the server, with the user voice characteristic information obtained by analyzing the voice signal of the user Register the user by storing user profile information, receive a user authentication request from an authentication requester directly or indirectly through the network, and search for a word or sentence corresponding to the user's status in a pre-built word group or sentence group. Selects and transmits to the authentication requestor a voice signal of the authentication requestor for the word or sentence or voice property information of the authentication requester extracted from the voice signal through a network, and receives the registered user voice property information. Using the voice signal Is a user authentication method using a network-based word group or sentence group in the server for determining whether the voice characteristic information of the authentication requester extracted from the voice signal is actually input by the user, corresponding to the method It is possible to provide an apparatus and a system.

According to another preferred aspect of the present invention, in network-based speaker learning for authenticating a user connected using a microphone at a server, a predetermined input including at least a member identifier is received from a client and stored in a website database. Determine a word group among a plurality of word groups pre-built in the web site database, where each word group consists of a plurality of words, and pre-built word group identifiers corresponding to the word groups. Combined with the input and stored in the web site database, a plurality of words belonging to the word group, a pre-built word identifier corresponding to each word, and a user corresponding to the client at the client pronouncing the words Preprocessing voice signals to each word Embed component software to extract the corresponding speaker components and send them to the server and send them to the client, and send word identifiers and speaker components corresponding to each word from the client. Receiving the speaker components, extracting reference pattern data corresponding to each word by performing a learning process, generating a reference pattern identifier corresponding to each reference pattern data, and then generating the member identifier and the reference pattern. A speaker learning method using a network-based word group for user authentication in a server storing data, the reference pattern identifier, and a word identifier in a voice information database, and an apparatus and system corresponding to the method may be provided.

In a preferred embodiment, the speaker learning method using a network-based word group for the user authentication in the server receives a connection request signal from the client, to input a predetermined input to the user (Form Transmitting a second document of the form to the client, receiving the predetermined input from the client, storing the predetermined document in the web site database, and determining whether there is a speaker component for which a learning process has not been performed. And if there is a speaker component for which the determination result learning process has not been performed, proceeding to extracting the reference pattern data, and if there is no speaker component for which the determination result learning process has not been performed, The method may further include transmitting a completion notification.

In another preferred embodiment, the predetermined input is a member identifier, name, gender, address, social security number, e-mail address. The determining of the one word group corresponds to at least one of a gender, an address, and an e-mail address.

According to still another aspect of the present invention, a speaker learning method using a network-based word group in a client corresponding to performing a network-based speaker learning method for authenticating a connected user using a microphone in a server is performed. A program of instructions that can be executed by a digital processing apparatus is tangibly embodied in order to provide a recording medium that can be read by a digital processing apparatus. The speaker learning method using a network-based word group in the client includes: Preprocessing words belonging to a specific word group from the server, a pre-built word identifier corresponding to each word, and a voice signal in which the user corresponding to the client pronounces the words. Add speaker components corresponding to words Receiving a first document embedding component software for transmission to the server, displaying the first document on a display device, and voice corresponding to each word from the user Receiving a signal, extracting a speaker component from a speech signal corresponding to each word and transmitting a word identifier and speaker component corresponding to each word to the server It is possible to provide a recording medium for user authentication in.

In a preferred embodiment, the speaker learning method using the network-based word group in the client transmits a connection request signal input by the user by the input means to the server, the user input from the server a predetermined input Receiving and displaying a second document of the form (Form) form that can be entered on the display device, Receiving the predetermined input from the user and transmitting to the server and Completed speaker learning from the server The method may further include receiving a notification and displaying the notification on the display device.

In another preferred embodiment, the predetermined input is a member identifier, name, gender, address, social security number, e-mail address.

In another preferred embodiment, the speaker learning method using a network-based word group in the client may further comprise the step of converting the voice signal to a digital signal (Analog to Digital Converte).

According to another aspect of the present invention, in a network-based speaker learning for authenticating a user connected using a microphone at a server, a predetermined input including at least a member identifier is received from a client and stored in a website database. Determine a word group among a plurality of word groups pre-built in the web site database, where each word group consists of a plurality of words, and pre-built word group identifiers corresponding to the word groups. Combined with the input, stored in the web site database, embeds a plurality of words belonging to the word group and a pre-built word identifier corresponding to each word in the first document and sends the same to the client. Word identifiers and voice signals corresponding to the respective words from the client Receiving, preprocessing the speech signals to extract speaker components corresponding to each word, extracting reference pattern data corresponding to each word by performing a learning process on the speaker components, and After generating a reference pattern identifier corresponding to each reference pattern data, a network-based word for user authentication in a server that stores the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. A speaker learning method using a group, an apparatus and a system corresponding to the method can be provided.

In a preferred embodiment, the speaker learning method using a network-based word group for the user authentication in the server receives a connection request signal from the client, to input a predetermined input to the user (Form Transmitting a second document of the form to the client, receiving the predetermined input from the client, storing the predetermined document in the web site database, and determining whether there is a speaker component for which a learning process has not been performed. And if there is a speaker component for which the determination result learning process has not been performed, proceeding to extracting the reference pattern data, and if there is no speaker component for which the determination result learning process has not been performed, The method may further include transmitting a completion notification.

In one preferred embodiment, the predetermined input is a member identifier, name, gender, address, social security number, e-mail address. The determining of the one word group corresponds to at least one of a gender, an address, and an e-mail address.

According to still another aspect of the present invention, a speaker learning method using a network-based word group in a client corresponding to performing a network-based speaker learning method for authenticating a connected user using a microphone in a server is performed. A program of instructions that can be executed by a digital processing apparatus is tangibly embodied in order to provide a recording medium that can be read by a digital processing apparatus. The speaker learning method using a network-based word group in the client includes: Receiving, from the server, a first document in which words belonging to a specific word group and a pre-built word identifier corresponding to each word are embedded, displaying the first document on a display device, a user Receiving a speech signal corresponding to each word from the and And transmitting a word identifier and voice signals corresponding to each word to the server.

In a preferred embodiment, the speaker learning method using the network-based word group in the client transmits a connection request signal input by the user by the input means to the server, the user input from the server a predetermined input Receiving and displaying a second document of the form (Form) form that can be entered on the display device, Receiving the predetermined input from the user and transmitting to the server and Completed speaker learning from the server The method may further include receiving a notification and displaying the notification on the display device.

In another preferred embodiment, the predetermined input is a member identifier, name, gender, address, social security number, e-mail address. Here, the speaker learning method using a network-based word group at the client may further include converting the voice signal into a digital signal.

According to still another aspect of the present invention, in a network-based speaker confirmation for authenticating a user connected using a microphone at a server, a member identifier is received from a client and stored in advance in a website database corresponding to the member identifier. Search for a word group identifier, select a word from among a plurality of words belonging to the word group corresponding to the word group identifier, the word, a word identifier corresponding to the word, and a voice signal in which the user pronounces the word Preprocessing and extracting a voice signal corresponding to the word and transmitting it to a server by embedding component software in a first document for transmission to a server, and transmitting the corresponding software word to the client. Receives an audio signal and receives the audio signal as an input. Extracting voice pattern data corresponding to a word identifier, searching for reference pattern data stored in a voice information database corresponding to the word identifier, determining whether the voice pattern data and the reference pattern data match, and determining the As a result, if the voice pattern data and the reference pattern data match, a permission notification is sent to the client. If the voice pattern data and the reference pattern data do not match, the server sending a reject notification to the client. A speaker identification method using a network-based word group for user authentication of an apparatus, and an apparatus and system corresponding to the method may be provided.

In a preferred embodiment, a speaker identification method using a network-based word group for user authentication at the server receives a connection request signal from the client, and forms a member identifier to the user. And sending the second document of to the client.

In another preferred embodiment, selecting one word is random selection.

In another preferred embodiment, selecting one word is selecting one of a plurality of words belonging to the word group in consideration of the time at the speaker identification, the speaker's state, and the connection location.

According to still another aspect of the present invention, a speaker identification method using a network-based word group in a client corresponding to performing a network-based speaker identification method for authenticating a connected user using a microphone in a server is performed. A program of instructions that can be executed by a digital processing apparatus is tangibly embodied in order to provide a recording medium that can be read by a digital processing apparatus, and the speaker identification method using a network-based word group in the client includes: Transmitting a connection request signal input by a user to the server to the server, receiving a second document in the form of a form from which the user can input a member identifier from the server to the display device; Displaying, input by the user by the input means Transmitting the member identifier to the server, preprocessing a word corresponding to the member identifier, a word identifier corresponding to the word, and a voice signal in which the user pronounces the word from the server to correspond to the word. Extracting a speaker component to receive a first document embedded with component software for transmitting to the server, displaying the first document on the display device, and the word from the user Receiving a speech signal corresponding to the speech signal, extracting a speaker component from the speech signal corresponding to the word, transmitting the speaker component to the server, and receiving a permission notification from the server. Displaying the permission notice and receiving a denial notice from the server, It is possible to provide a recording medium for the user authentication in a client comprising the step of displaying the rejection notice to the group display.

In a preferred embodiment, the speaker identification method using a network-based word group in the client may further comprise the step of converting the voice signal to a digital signal (Analog to Digital Converte).

According to still another aspect of the present invention, in a network-based speaker identification for authenticating a user connected using a microphone at a server, a member identifier is received from a client and a word is stored in a website database corresponding to the member identifier. Search for a group identifier, select one word among a plurality of words belonging to the word group corresponding to the word group identifier, embed the word and the word identifier corresponding to the word in the first document, and Transmit to a client, receive a voice signal corresponding to the word from the client, preprocess the voice signal to extract the speaker component corresponding to the word, and input the speaker component as the input to correspond to the word Extracts speech pattern data, and corresponds to the word identifier Search for the reference pattern data stored in the information database, determine whether the voice pattern data and the reference pattern data match, and if the voice pattern data and the reference pattern data match, the permission notification to the client If the voice pattern data and the reference pattern data does not match, the speaker identification method using a network-based word group for user authentication at the server for transmitting a rejection notification to the client, Corresponding devices and systems can be provided.

In a preferred embodiment, a speaker identification method using a network-based word group for user authentication at the server receives a connection request signal from the client, and forms a member identifier to the user. The method may further include transmitting a second document of the client to the client.

In another preferred embodiment, selecting one word is random selection.

In another preferred embodiment, selecting one word is selecting one of a plurality of words belonging to the word group in consideration of the time at the speaker identification, the speaker's state, and the connection location.

According to still another aspect of the present invention, a speaker identification method using a network-based word group in a client corresponding to performing a network-based speaker identification method for authenticating a connected user using a microphone in a server is performed. A program of instructions that can be executed by a digital processing apparatus is tangibly embodied in order to provide a recording medium that can be read by a digital processing apparatus. Transmitting a connection request signal input by a user to the server to the server, receiving a second document in the form of a form from which the user can input a member identifier from the server to the display device; Displaying by the user by the input means. Transmitting the displayed member identifier to the server, receiving a word corresponding to the member identifier and a first document including the word identifier corresponding to the word from the server, and displaying the first document. Displaying on the device, receiving a voice signal corresponding to the word from the user, transmitting the voice signal to the server, and when receiving a permission notification from the server, sending the permission notification to the display device. And displaying the rejection notice on the display device when the rejection notification is received from the server.

In a preferred embodiment, the speaker identification method using a network-based word group in the client may further comprise the step of converting the voice signal to a digital signal (Analog to Digital Converte).

According to another aspect of the present invention, in network-based speaker learning for authenticating a user connected by using a microphone of a network server system among server systems, receiving a predetermined input including at least a member identifier from a client A word group of a plurality of word groups, wherein each word group consists of a plurality of words, and determines a dictionary corresponding to the word group. Combines the constructed word group identifiers with the inputs and stores them in the web site database, the plurality of words belonging to the word group, a pre-built word identifier corresponding to each word, and the user at the client Preprocessing the spoken voice signal to correspond to each word Embedding component software in a first document for extracting speaker components to be transmitted to the server and transmitting the same to the client, and sending a member identifier, a word identifier corresponding to each word from the client, Receiving a speaker component corresponding to each word and transmitting them to a speaker recognition server system, receiving a learning completion notification from the speaker recognition server system for transmitting to the client for user authentication at the network server system A speaker learning method using a network-based word group, an apparatus and a system corresponding to the method can be provided.

In a preferred embodiment, the speaker learning method using a network-based word group for user authentication in a network server system of the server system receives a connection request signal from the client, to input a predetermined input to the user The method may further include transmitting a second document in a form form to the client, and receiving the predetermined input from the client and storing the input document in the web site database.

In another preferred embodiment, the predetermined input is a member identifier, name, gender, address, social security number, e-mail address. The determining of the one word group corresponds to at least one of a gender, an address, and an e-mail address.

According to another aspect of the present invention, in network-based speaker learning for authenticating a user connected using a microphone in a speaker recognition system among server systems, a member identifier input directly from a client or via a network server system Receiving word identifiers and speaker components corresponding to each word, extracting reference pattern data corresponding to each word by performing a learning process on the speaker components, and a reference corresponding to each reference pattern data After generating a pattern identifier, a network-based word group for user authentication in a speaker recognition server system among server systems storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker learning method, equivalent to the above method It can provide devices and systems.

In a preferred embodiment, the speaker learning method using a network-based word group for user authentication in the speaker recognition server system of the server system determines whether there is a speaker component for which the learning process has not been performed, If there is a speaker component for which the learning process has not been performed as a result of the determination, the method proceeds to the step of extracting the reference pattern data; Or transmitting to the network server system.

According to another aspect of the present invention, in a client corresponding to performing a network-based speaker learning method for authenticating a connected user using a microphone in a network server system and a speaker recognition server system of the server system In order to perform the speaker learning method using the network-based word group of the program of instructions that can be executed by the digital processing device is tangibly implemented, and in the recording medium that can be read by the digital processing device, in the client The speaker learning method using the network-based word group of the system includes words belonging to a specific word group from the network server system, a pre-built word identifier corresponding to each word, and a voice of the user pronounced words in the client. Preprocess the signal receiving a first document by embedding component software for extracting speaker components corresponding to each word and transmitting the same to a server; and displaying the first document on a display device. Receiving a speech signal corresponding to each word from the user, extracting a speaker component from the speech signal corresponding to each word and the member identifier, word identifiers corresponding to each word And transmitting the speaker components to the network server system or the speaker recognition server system.

In a preferred embodiment, the speaker learning method using a network-based word group in the client transmits a connection request signal input by the user by an input means to the network server system, the user from the network server system Receiving and displaying a second document in a form of a form capable of inputting a predetermined input item on the display device, receiving the predetermined input item from the user, and transmitting it to the network server system; and The method may further include receiving a speaker learning completion notification from the network server system or the speaker recognition server system and displaying the notification on the display device.

In another preferred embodiment, the predetermined input is a member identifier, name, gender, address, social security number, e-mail address.

In another preferred embodiment, the speaker learning method using a network-based word group in the client may further comprise the step of converting the voice signal to a digital signal (Analog to Digital Converte).

According to another aspect of the present invention, in network-based speaker learning for authenticating a user connected by using a microphone of a network server system among server systems, receiving a predetermined input including at least a member identifier from a client Determine a word group among a plurality of word groups, wherein each word group consists of a plurality of words, and store a dictionary in a web site database and construct a dictionary corresponding to the word group. Combined word group identifiers with the inputs and stored in the web site database, embedding a plurality of words belonging to the word group and a pre-built word identifier corresponding to each word in the first document. Send to the client, and from the client a member identifier, the respective A network server system of the server system which receives a word identifier corresponding to a language and a voice signal corresponding to each word and transmits it to a speaker recognition server system, and receives a learning completion notification from the speaker recognition server system and transmits it to the client. A speaker learning method using a network-based word group for user authentication in, and an apparatus and system corresponding to the method can be provided.

In a preferred embodiment, the speaker learning method using a network-based word group for user authentication in a network server system of the server system receives a connection request signal from the client, to input a predetermined input to the user The method may further include transmitting a second document in a form form to the client, and receiving the predetermined input from the client and storing the input document in the web site database.

In another preferred embodiment, the predetermined input is a member identifier, name, gender, address, social security number, e-mail address. The determining of the one word group corresponds to at least one of a gender, an address, and an e-mail address.

According to another aspect of the present invention, in network-based speaker learning for authenticating a user connected using a microphone in a speaker recognition system among server systems, a member identifier input directly from a client or via a network server system Receiving word identifiers and voice signals corresponding to each word, preprocessing the voice signals to extract speaker components corresponding to each word, and performing the learning process on the speaker components. After extracting reference pattern data corresponding to a word, generating a reference pattern identifier corresponding to each reference pattern data, and storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Server system of the speaker recognition server system The speaker learning method using the network-based group of words for user authentication, it is possible to provide an apparatus and a system corresponding to the method.

In a preferred embodiment, the speaker learning method using a network-based word group for user authentication in the speaker recognition server system of the server system determines whether there is a speaker component for which the learning process has not been performed, If there is a speaker component for which the learning process has not been performed as a result of the determination, the method proceeds to the step of extracting the reference pattern data; and if there is no speaker component for which the learning process has not been performed, the client is notified of completion. Or transmitting to the network server system.

According to another aspect of the present invention, the network-based at the client corresponding to performing a network-based speaker learning method for authenticating the connected user using a microphone in the network server system and the speaker recognition server system of the server system A program of instructions that can be executed by a digital processing apparatus is tangibly embodied to perform a speaker learning method using the word group of. In a recording medium that can be read by a digital processing apparatus, the network-based method of the client is performed. A speaker learning method using a word group of a, receiving a first document in which the words belonging to a specific word group and a pre-built word identifier corresponding to each word from the network server system (Embedding), Displaying the first document on a display device; Receiving a voice signal corresponding to each word from a user and transmitting the member identifier, word identifiers corresponding to each word, voice signals to the network server system or the speaker recognition server system; A recording medium for user authentication in a client can be provided.

In a preferred embodiment, the speaker learning method using a network-based word group in the client transmits a connection request signal input by the user by an input means to the network server system, the user from the network server system Receiving and displaying a second document in a form of a form capable of inputting a predetermined input item on the display device, receiving the predetermined input item from the user, and transmitting it to the network server system; and The method may further include receiving a speaker learning completion notification from the network server system or the speaker recognition server system and displaying the notification on the display device.

In another preferred embodiment, the predetermined input is a member identifier, name, gender, address, social security number, e-mail address.

In another preferred embodiment, the speaker learning method using a network-based word group in the client may further comprise the step of converting the voice signal to a digital signal (Analog to Digital Converte).

According to still another aspect of the present invention, in a network-based speaker confirmation for authenticating a user connected using a microphone in a network server system among server systems, a member identifier is received from a client and corresponds to the member identifier. Search for a word group identifier stored in a web site database, select a word among a plurality of words belonging to the word group, and pronounce the word by the user at the word, the word identifier corresponding to the word, and the client Embedding component software into a first document for preprocessing a voice signal, extracting a speaker component corresponding to the word, and transmitting it to the network server system, and transmitting the same to the client; Receiving the speaker component from the Sends a child component, the member identifier and the word identifier to a speaker recognition server system, receives a permission notification from the speaker recognition server system and sends it to the client, receives a rejection notification from the speaker recognition server system, and A speaker identification method using a network-based word group for user authentication in a network server system among a server system transmitted to a client, and an apparatus and system corresponding to the method may be provided.

In a preferred embodiment, a speaker identification method using a network-based word group for user authentication in a network server system of the server system may receive a connection request signal from the client to input a member identifier to the user The method may further include transmitting a second document in a form form to the client.

In another preferred embodiment, selecting one word is random selection.

In another preferred embodiment, selecting one word is selecting one of a plurality of words belonging to the word group in consideration of the time at the speaker identification, the speaker's state, and the connection location.

In another embodiment of the present invention, in a network-based speaker identification for authenticating a user connected using a microphone in a speaker recognition system of a server system, a member identifier input directly from a client or via a network server system Receive the word identifier and the speaker component, extract the voice pattern data corresponding to the word identifier by inputting the speaker component, retrieve the reference pattern data stored in the voice information database corresponding to the word identifier, It is determined whether the pattern data and the reference pattern data match. If the voice pattern data and the reference pattern data match, a permission notification is sent to the client or the network server system. Having said voice pattern Speaker and a method of identifying a speaker using a network-based word group for user authentication in a speaker recognition server system among a server system transmitting a rejection notification to the client or the network server system when the reference pattern data does not match. It is possible to provide an apparatus and system corresponding to the method.

According to another aspect of the present invention, the network-based at the client corresponding to performing the network-based speaker identification method for authenticating the connected user by using the microphone of the network server system and the speaker recognition server system of the server system A program of instructions that can be executed by a digital processing apparatus is tangibly embodied to perform a speaker identification method using the word group of. A recording medium that can be read by a digital processing apparatus, the network-based in the client A speaker identification method using a group of words, the method comprising: transmitting a connection request signal input by an input means by an input means to the network server system, and a member identifier can be input from the network server system to the user. Form of the second document Sending a member identifier input by the user to the network server system to the network server system; a word corresponding to the member identifier from the network server system; and a word identifier corresponding to the word. And inserting component software for preprocessing a voice signal in which the user pronounces the word, extracting a speaker component corresponding to the word, and transmitting the extracted speaker component to the network server system or the speaker recognition server system. Receiving a first document embedding, displaying the first document on the display device, receiving a voice signal corresponding to the word from the user, and extracting a speaker component from the voice signal corresponding to the word. Extracting the speaker component from the network; Transmitting to the system or the speaker recognition server system; when receiving the permission notification from the network server system or the speaker recognition server system, displaying the permission notification on the display device; and the network server system or the speaker recognition. If receiving a rejection notification from a server system, displaying the rejection notification on the display device.

In a preferred embodiment, the speaker identification method using a network-based word group in the client may further comprise the step of converting the voice signal to a digital signal (Analog to Digital Converte).

According to still another aspect of the present invention, in a network-based speaker confirmation for authenticating a user connected using a microphone in a network server system among server systems, a member identifier is received from a client and corresponds to the member identifier. Search for a word group identifier stored in a web site database, select one word among a plurality of words belonging to the word group corresponding to the word group identifier, and convert the word and the word identifier corresponding to the word into the first document. Embed and transmit to the client, receive a voice signal from the client, transmit the voice signal, the member identifier and the word identifier to a speaker recognition server system, and receive a permission notification from the speaker recognition server system. Send it to the client, To provide a speaker identification method using a network-based word group for user authentication in a network server system among the server systems receiving a rejection notification from a speaker recognition server system and transmitting the same to the client, an apparatus and a system corresponding thereto. Can be.

In a preferred embodiment, a speaker identification method using a network-based word group for user authentication in a network server system of the server system may receive a connection request signal from the client to input a member identifier to the user The method may further include transmitting a second document in a form form to the client.

In another preferred embodiment, selecting one word is random selection.

In another preferred embodiment, selecting one word is selecting one of a plurality of words belonging to the word group in consideration of the time at the speaker identification, the speaker's state, and the connection location.

According to another aspect of the present invention, a member identifier input from a client or via a network server system in network-based speaker identification for authenticating a user connected using a microphone in a speaker recognition system among server systems Receiving a word identifier and a speech signal, preprocessing the speech signal to extract a speaker component corresponding to the word, extracting the speech pattern data corresponding to the word by inputting the speaker component, Search for reference pattern data stored in a voice information database corresponding to a word identifier, determine whether the voice pattern data and the reference pattern data match, and if the voice pattern data and the reference pattern data match, , The client or the net A user at the speaker recognition server system of the server system which transmits a permission notification to the network server system and transmits a rejection notification to the client or the network server system if the voice pattern data and the reference pattern data do not match. A speaker identification method using a network-based word group for authentication, and an apparatus and system corresponding to the method can be provided.

According to another aspect of the present invention, the network-based at the client corresponding to performing the network-based speaker identification method for authenticating the connected user by using the microphone of the network server system and the speaker recognition server system of the server system A program of instructions that can be executed by a digital processing apparatus is tangibly embodied to perform a speaker identification method using the word group of. A recording medium that can be read by a digital processing apparatus, the network-based in the client A speaker identification method using a group of words, the method comprising: transmitting a connection request signal input by an input means by an input means to the network server system, and a member identifier can be input from the network server system to the user. Form of the second document Sending a member identifier input by the user by the input means to the network server system; a word corresponding to the member identifier and a word corresponding to the word from the network server system Receiving a first document embedded with an identifier, displaying the first document on the display device, receiving a voice signal corresponding to the word from the user, and transmitting the voice signal to the network server. Transmitting to the system or the speaker recognition server system; when receiving the permission notification from the network server system or the speaker recognition server system, displaying the permission notification on the display device; and the network server system or the speaker recognition. Rejection notice from server system And receiving the reject notification on the display device.

In a preferred embodiment, the method of speaker identification using a network-based word group in the client further comprises the step of converting the voice signal to a digital signal (Analog to Digital Converte). According to another aspect of the present invention In the e-commerce, after the customer makes at least one order or subscription, and the contract is established, the voice information restoration method for preventing the customer from making or denying the fact of the contract establishment or ordering, each of the prestored word groups Transmitting a word or sentence corresponding to the order or subscription of the client to the client, receiving a purchase order voice signal corresponding to the word or sentence from the client, and receiving voice pattern data corresponding to the purchase order voice signal. And storing the data in a database. May provide a method for restoring voice information using a word group. In a preferred embodiment, the method for restoring voice information using the word group may include the voice pattern data when the customer establishes or denies a fact of a contract or an order. The method may further include reconstructing the purchase order voice signal. In another preferred embodiment, the voice information reconstruction method using the word group may include: receiving the purchase order voice signal from a client; The method may further include extracting a speaker component corresponding to the purchase order voice signal by preprocessing a signal, and extracting voice pattern data corresponding to the purchase order voice signal from the speaker component. In another embodiment, the method of restoring voice information using the word group may be large. The method may further include receiving a speaker component preprocessing the purchase order voice signal from an client and extracting the voice pattern data corresponding to the purchase order voice signal from the speaker component. In an embodiment, the voice information restoration method using the word group may include receiving the purchase order voice signal from a client, preprocessing the purchase order voice signal, and extracting a speaker component corresponding to the purchase order voice signal. The method may further include extracting voice pattern data corresponding to the purchase order voice signal from the speaker component and outputting the restored purchase order voice signal to the Gauguin output device. In an embodiment, the voice information restoration room using the word group Receiving a speaker component preprocessing the purchase order voice signal from a client, extracting the voice pattern data corresponding to the purchase order voice signal from the speaker component and the restored purchase order voice signal The method may further include outputting to the output device of the customer.

Next, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 shows a schematic configuration diagram of a device to which the present invention can be applied.

Referring to FIG. 1, a network-based speaker learning and speaker identification apparatus according to an exemplary embodiment of the present invention may include a client 103 which extracts a speaker component which is a speech signal or a voice feature input from a user, and a general network. A server 101 that not only performs services, but also performs speaker identification, speaker recognition and management coordination, a web site database 105 that stores personal information and general site information, and a voice database that stores voice information of members. 107.

Figure 2a shows another schematic diagram of a device to which the present invention can be applied.

2b shows another schematic configuration diagram of a device to which the present invention can be applied.

2A and 2B, a network-based speaker learning and speaker identification apparatus according to another preferred embodiment of the present invention is a client 209a which extracts a speaker component which is a speech signal or a voice feature element that is schematically input from a user. 209b), network servers 201a and 201b that perform general network services, speaker identification tasks, speaker recognition tasks, and other management coordinating speaker identification servers 203a and 203b, storing member's personal information and general site information. Web site databases 205a and 205b and voice databases 207a and 207b that store voice information of members.

Comparing the apparatus illustrated in FIG. 1 with the apparatus illustrated in FIGS. 2A and 2B, the server 101 in FIG. 1 is connected to the network servers 201a and 201b and the speaker recognition server 203a in FIGS. 2A and 2B. Different from 203b). The present invention may be performed integrally as shown in FIG. 1 by hardware or software, or may be separately performed as shown in FIG. Accordingly, embodiments of the present invention will be described in the future by the apparatus as illustrated in FIG. 1 and the apparatus as illustrated in FIG. 2B.

First, an embodiment of the present invention will be described by the apparatus as illustrated in FIG. 1.

3 is a flowchart illustrating a speaker learning method according to an exemplary embodiment of the present invention.

When the user transmits a connection request signal from the client 103 via an input device (step 301), the server 101 receives the connection request signal (step 303) and sends a first HTML document to the client 103. Is transmitted (step 305).

Here, the first HTML document may be a user input a predetermined input, for example, a member identifier, name, gender, address, social security number, e-mail address, etc. in the HTML document for performing a general network service Embedding an HTML document in the form of a form.

The client 103 receives the first HTML document and displays it on the display device (step 307). Thereafter, the user inputs the predetermined input item, for example, a member identifier, a name, a gender, an address, a social security number, an e-mail address, and the like through the input device, and transmits it to the server 101 (step 309). )do.

The server 101 receives the predetermined input and stores it in the web site database 105 (step 311). According to a preferred embodiment of the present invention, the web site database 105 includes at least a member table, a word group table and a word table, wherein the member table includes a member identifier field, a name field, a social security number field and a gender field. There is an address field, an e-mail address field and a word family identifier field (not shown). The word group table includes a word group identifier field and a plurality of word identifier fields. The word table includes a word identifier field and a word field.

The speaker recognition system in which the speaker has a fixed sentence to be spoken is called text dependent, and the case where the sentence is undefined and pronounced freely is called text independent. Character dependent type has a disadvantage of easily deceiving a speaker recognition system using previously recorded voice information. In order to make up for this drawback, the method and apparatus for network-based speaker learning and speaker science of the present invention are based on a text prompted speaker recognition method.

Thereafter, the server 101 can learn from a plurality of word group tables that have been previously constructed using the predetermined input, for example, a member identifier, name, gender, address, social security number, e-mail address, and the like. The word group is determined (step 313). Each of the plurality of pre-built word groups is composed of a plurality of words. In the determination method, for example, if a user's address is Seoul, a word group to which a word related to Seoul belongs is determined among a plurality of pre-built word groups. Thus, determining the word group to learn preferably corresponds to at least one of the predetermined inputs, eg, gender, address, e-mail address.

The server 101 then transmits a second HTML document to the client 103 (step 315). The second HTML document may preprocess the words belonging to the word group to be learned determined by the predetermined input, a pre-built word identifier corresponding thereto, and a voice signal in which the user pronounces the words in the client 103 ( By embedding component software for preprocessing to extract speaker components corresponding to each word and transmit them to the server.

The client 103 receives the second HTML document and displays it on the display device (317), and according to a preferred embodiment of the present invention, receives a voice signal corresponding to each word from the user through a microphone (319) )do. It is desirable for the user to speak four or more times for each speaker. Thereafter, the client 103 performs an A / D conversion (not shown) of the voice signal and then performs a preprocessing process using the same as an input.

4 is a schematic flowchart illustrating a preprocessing process according to an exemplary embodiment of the present invention.

In the preprocessing, it is a question of which feature feature, that is, the speaker component, is selected. Voice signals contain a mixture of information and noise. This includes information about the speaker's identity, as well as other information. From the point of view of speaker identification, all other information besides the speaker's identity is a kind of noise. This noise causes a decrease in recognition rate. Therefore, all information other than the speaker's identity should be extracted during the preprocessing.

Referring to FIG. 4, a pretreatment process according to an exemplary embodiment of the present invention may include removing a direct current component (step 401), an energy normalization step (step 403), a silent section removing step (step 405), and a feature element. Extracting the speaker component (step 407).

The client 103 preprocesses the A / D converted voice signal to extract a speaker component (step 321). The member identifier, the word identifier and the speaker component are then sent to the server 101 (step 323).

5 is a diagram illustrating a distributed processing scheme according to a preferred embodiment of the present invention.

Referring to FIG. 5, the speaker learning system according to the preferred embodiment of the present invention is largely composed of a client component 503 and a server component 501. The client component 503 is composed of two interfaces. It consists of an interface that extracts the speaker component from the digitized voice signal and an interface that is informed of the recognition result from the server. The proxy object performs a marshalling process to packetize the extracted speaker component and then transmits it to the server through a channel and receives voice pattern data from the packet transmitted to the server through a stub object. Perform unmarshalling process to restore and deliver speaker component to recognition process interface in server's recognition component. After performing the recognition process, the result is notified to the result operation interface of the client in the reverse order of the previous procedure.

By such a distributed processing scheme, even if many users access the speaker verification system on the Internet at the same time, it is possible to prevent the load on the entire system from being concentrated on some systems.

The server 101 receives (step 325) the words, word identifiers, and speaker components from the client. Thereafter, reference pattern data is extracted through performing a learning process on the speaker components (step 327). The reference pattern data is a value used as a reference for comparison when speaker identification is required.

6 is a flowchart schematically illustrating the learning process according to an embodiment of the present invention.

Referring to FIG. 6, the learning process includes a K-means crusting step (step 601), a vector quantization step (step 603), a probability value initialization step (step 605), a forward backward algorithm step (step 607), and Baum-welch estimation. Step 609 is included.

The server 101 stores the extracted reference pattern data in the voice information database 107 (step 329). According to a preferred embodiment of the present invention, the voice information database 107 has at least a reference pattern data table, a member-reference pattern identifier table and a word table (not shown). The reference pattern data table includes a reference pattern identifier field and a reference pattern data field, the member-reference pattern identifier table includes a member identifier field and a plurality of reference pattern identifier fields, and the word table includes a word identifier field and a reference pattern identifier. There is a field.

Thereafter, the server 101 determines whether there is a speaker component for which the learning process has not been performed (step 331), and if the determination result is positive, the server 101 returns to the reference pattern data extraction step (step 327), If the determination is negative, a completion notification is sent to the client 103 (step 333).

The client receives the completion notification from the server 101 and displays it on the display device (step 335).

7 is a flowchart illustrating a speaker learning method according to another exemplary embodiment of the present invention.

Referring to FIG. 7, the speaker learning method is generally the same as the speaker learning method as shown in FIG. Thus, a similar description may be substituted.

When the user transmits a connection request signal from the client 103 through an input device (step 701), the server 101 receives the connection request signal (step 703) and sends a third HTML document to the client 103. Is transmitted (step 705).

Here, the third HTML document may allow a user to input predetermined input items, for example, a member identifier, name, gender, address, social security number, e-mail address, etc., into an HTML document for performing a general network service. Embedding an HTML document in the form of a form.

The client 103 receives the third HTML document and displays it on the display device (step 707). Thereafter, the user inputs the predetermined input item, for example, a member identifier, a name, a gender, an address, a social security number, an e-mail address, and the like through the input device, and transmits it to the server 101 (step 709). )do.

The server 101 receives the predetermined input and stores it in the web site database 105 (step 711). According to a preferred embodiment of the present invention, the web site database 105 includes at least a member table, a word group table and a word table, wherein the member table includes a member identifier field, a name field, a social security number field and a name field. There is an address field, an e-mail address field and a word family identifier field (not shown). The word group table includes a word group identifier field and a plurality of word identifier fields. The word table includes a word identifier field and a word field.

Thereafter, the server 101 determines a word group to be learned from a plurality of word group tables previously constructed using the predetermined input items, for example, sex, name, e-mail address, and the like (step 713). . Each of the plurality of pre-built word groups is composed of a plurality of words. In the determination method, for example, if a user's address is Seoul, a word group to which a word related to Seoul belongs is determined among a plurality of pre-built word groups. Thus, determining the word group to learn preferably corresponds to at least one of the predetermined inputs, eg, gender, full name, e-mail address.

The server 101 then transmits a fourth HTML document to the client 103 (step 715). The fourth HTML document embeds words belonging to the word group to be learned determined by the predetermined input item and corresponding pre-built word identifiers.

The client 103 receives the fourth HTML document and displays it on the display device (step 717). According to another preferred embodiment of the present invention, the client 103 receives a voice signal corresponding to each word from a user through a microphone. Step 719). It is desirable for the user to speak four or more times for each speaker. The client 103 then A / D converts the voice signal (not shown) to transmit the word identifier and the voice signal to the server 101 (step 721).

The server 101 receives the word identifier and the voice signal from the client 103 (step 723) and extracts the speaker component through the preprocessing process (step 725). Thereafter, reference speaker data is extracted through performing the learning process on the speaker components (step 727).

The server 101 stores the extracted reference pattern data in the voice information database 107 (step 729). According to a preferred embodiment of the present invention, the voice information database 107 has at least a reference pattern data table, a member-reference pattern identifier table and a word table (not shown). The reference pattern data table includes a reference pattern identifier field and a reference pattern data field, the member-reference pattern identifier table includes a member identifier field and a plurality of reference pattern identifier fields, and the word table includes a word identifier field and a reference pattern identifier. There is a field.

Thereafter, the server 101 determines whether there is a speaker component for which the learning process has not been performed (step 731), and if the determination result is affirmative, the server 101 returns to the reference pattern data extraction step (step 727), If the determination is negative, a completion notification is sent to the client 103 (step 733).

The client receives the completion notification from the server 101 and displays it on the display device ( step 735).

8 is a flowchart illustrating a speaker identification method according to an embodiment of the present invention.

When the user transmits a connection request signal from the client 103 through an input device (step 801), the server 101 receives the connection request signal (step 803) and sends a fifth HTML document to the client 103. Is transmitted (step 805).

Here, the fifth HTML document is an embedded HTML document in which a user can input a member identifier into an HTML document for performing a general network service.

The client 103 receives the fifth HTML document and displays it on the display device (step 807). Thereafter, the member identifier is received from the user through the input device and transmitted to the server 101 (step 809).

The server 101 receives the member identifier (step 811) and retrieves the word group identifier stored in the web site database 105 corresponding to the member identifier (step 813).

The word group is built in a dictionary, and the server 101 randomly selects one word among a plurality of words belonging to the word group (step 815). However, each time the speaker is identified, one of a plurality of words belonging to the word group may be selected in consideration of the time, the speaker's state, and the connection location. For example, when the user is in a nasal cold state, the user selects the remainder except for the non-negative form (for example, the word 'mom' includes the non-negative).

A speaker recognition system in which a sentence to be uttered by a speaker is fixed is referred to as text dependent, and a case in which a sentence is undefined and pronounced freely is referred to as text independent. Character dependent type has a disadvantage of easily deceiving a speaker recognition system using previously recorded voice information. In order to compensate for this drawback, the method and apparatus for network-based speaker learning and speaker studies of the present invention are based on a text prompted speaker recognition method.

The server 101 then sends a sixth HTML document to the client 103 (step 817). The sixth HTML document preprocesses a pre-established word identifier corresponding to the word and a voice signal in which the user pronounces the word in the client 103 to extract a speaker component corresponding to the word to the server. Embedding component software for transmission.

The client 103 receives the sixth HTML document and displays it on the display device (step 819). According to an exemplary embodiment of the present invention, the client 103 receives a voice signal corresponding to each word from a user through a microphone. Step 821). Thereafter, the client 103 performs an A / D conversion (not shown) of the voice signal and then performs a preprocessing process using the same as an input.

9 is a schematic flowchart illustrating a speaker identification process according to another exemplary embodiment of the present invention.

Referring to FIG. 9, the process extracts the speech pattern data through the VQ indexing step (step 901) and the observation parameter extraction step (step 903) using the speaker component as an input, and in the Viterbi calculation step (step 905). And comparing the voice pattern data with the corresponding reference pattern data to determine the result (step 907).

In the preprocessing, it is a question of which feature feature, that is, the speaker component, is selected. Voice signals contain a mixture of information and noise. This includes information about the speaker's identity, as well as other information. From the point of view of speaker identification, all other information besides the speaker's identity is a kind of noise. This noise causes a decrease in recognition rate. Therefore, all information other than the speaker's identity should be extracted during the preprocessing.

The speaker component is again extracted as the speaker component through a preprocessing process as shown in FIG. 4.

The client 103 preprocesses the A / D converted voice signal to extract a speaker component (step 823). Thereafter, the speaker component is transmitted to the server 101 (step 825).

5 is a diagram illustrating a distributed processing scheme according to a preferred embodiment of the present invention.

Referring to FIG. 5, the speaker learning system according to the preferred embodiment of the present invention is largely composed of a client component 503 and a server component 501. The client component 503 is composed of two interfaces. It consists of an interface that extracts the speaker component from the digitized voice signal and an interface that is informed of the recognition result from the server. The preprocessing process interface delivers the extracted speaker component to the proxy object. The proxy object performs a marshalling process to packetize the extracted speaker component, transmits it to the server through a channel, and restores the speaker component from the packet transmitted to the server through a stub object. Unmarshalling is performed to deliver the speaker component to the recognition process interface in the server's recognition component. After performing the recognition process, the result is notified to the result operation interface of the client in the reverse order of the previous procedure.

By such a distributed processing scheme, even if many users access the speaker verification system on the Internet at the same time, it is possible to prevent the load on the entire system from being concentrated on some systems.

The server 101 receives (step 827) the speaker component from the client. The server 101 then extracts the speech pattern data corresponding to the word identifier (step 828) in the manner shown in FIG. Thereafter, reference pattern data stored in the voice information database 107 corresponding to the word identifier and the voice pattern data is retrieved (step 829).

Thereafter, the server 101 determines whether the voice pattern data and the reference pattern data match (step 831), and if the determination is affirmative, transmits a permission notification to the client 103 (step 833). If the determination is negative, a rejection notification is sent to the client 103 (step 835).

The client receives the completion notification from the server 101 and displays it on the display device (step 837) or receives the rejection notification and displays it on the display device (step 839).

10 is a flowchart illustrating a speaker identification method according to another exemplary embodiment of the present invention.

Referring to FIG. 10, the speaker learning method is generally the same as the speaker learning method as shown in FIG. Thus, a similar description may be substituted.

When the user transmits a connection request signal from the client 103 through an input device (step 1001), the server 101 receives the connection request signal (step 1003) and sends the seventh HTML document to the client 103. (Step 1005).

Here, the seventh HTML document is an embedded HTML document in which a user can input a member identifier into an HTML document for performing a general network service.

The client 103 receives the fifth HTML document and displays it on the display device (step 1007). Thereafter, a member identifier is input from the user through the input device and transmitted to the server 101 (step 1009).

The server 101 receives the member identifier (step 1011) and retrieves the word group identifier stored in the web site database 105 corresponding to the member identifier (step 1013).

The word group is built in a dictionary, and the server 101 arbitrarily selects one word among a plurality of words belonging to the word group (step 1015). However, each time the speaker is identified, one of a plurality of words belonging to the word group may be selected in consideration of the time, the speaker's state, and the connection location. For example, when the user is in a nasal cold state, the user selects the remainder except for the non-negative form (for example, the word 'mom' includes the non-negative).

The speaker recognition system in which the speaker has a fixed sentence to be spoken is called text dependent, and the case where the sentence is undefined and pronounced freely is called text independent. Character dependent type has a disadvantage of easily deceiving a speaker recognition system using previously recorded voice information. In order to make up for this drawback, the method and apparatus for network-based speaker learning and speaker science of the present invention are based on a text prompted speaker recognition method.

The server 101 then transmits an eighth HTML document to the client 103 (step 1017). The sixth HTML document is embedded with a pre-built word identifier corresponding to the word.

The client 103 receives the eighth HTML document and displays it on a display device (step 1019), and according to a preferred embodiment of the present invention, receives a voice signal corresponding to each word from a user through a microphone ( Step 1021). The client 103 then A / D converts the voice signal (not shown) and sends it to the server (step 1023).

The server 101 receives the voice signal from the client (step 1025). Thereafter, voice pattern data corresponding to the word identifier is extracted through the process as shown in FIG. 9 (step 1027).

9 is a schematic flowchart illustrating a speaker identification process according to an exemplary embodiment of the present invention.

Referring to FIG. 9, the process extracts the speech pattern data through the VQ indexing step (step 901) and the observation parameter extraction step (step 903) using the speaker component as an input, and in the Viterbi calculation step (step 905). And comparing the voice pattern data with the corresponding reference pattern data to determine the result (step 907).

Thereafter, the server 101 searches (step 1029) reference pattern data stored in the voice information database 107 corresponding to the word identifier and the voice pattern data.

Thereafter, the server 101 determines whether the voice pattern data and the reference pattern data match (step 1031), and if the determination is affirmative, transmits a permission notification to the client 103 (step 1033). If the result of the determination is negative, a rejection notification is sent to the client 103 (step 1035).

The client receives the completion notification from the server 101 and displays it on the display device (step 1037) or receives the rejection notification and displays it on the display device (step 1039).

Another embodiment of the present invention is described by the apparatus as illustrated in FIG. 2B.

11 is a flowchart illustrating a speaker learning method according to another preferred embodiment of the present invention.

When the user transmits a connection request signal from the client 209b through an input device (step 1101), the network server system 201b receives the connection request signal (step 1103), and sends the ninth request to the client 209b. The HTML document is sent (step 1105).

Here, the ninth HTML document may allow a user to input predetermined input items, for example, a member identifier, a name, a gender, an address, a social security number, an e-mail address, and the like into an HTML document for performing a general network service. Embedding an HTML document in the form of a form.

The client 209b receives the ninth HTML document and displays it on the display device (step 1107). Thereafter, the user inputs the predetermined input items, for example, a member identifier, a name, a gender, an address, a social security number, an e-mail address, and the like through the input device and transmits the same to the network server system 201b. Step 1109).

The network server system 201b receives the predetermined input and stores it in the web site database 105 (step 1111). The data structure of the web site database 205b is as described above.

Thereafter, the network server system 201b determines a word group to learn from a plurality of word group tables constructed in advance using the predetermined input items, for example, gender, address, and e-mail address (step 1113). )do. Each of the plurality of pre-built word groups is composed of a plurality of words. The word group determination method is as described above.

The network server system 201b then transmits a tenth HTML document to the client 209b (step 1115). The tenth HTML document may preprocess the words belonging to the word group to be learned determined by the predetermined input, a pre-built word identifier corresponding thereto, and a voice signal in which the user pronounces the words in the client 209b. By embedding component software for preprocessing to extract speaker components corresponding to each word and transmit them to the server.

The client 209b receives the tenth HTML document and displays it on the display device (step 1117). According to an exemplary embodiment of the present invention, the client 209b receives a voice signal corresponding to each word from a user through a microphone. Step 1119). It is desirable for the user to speak four or more times for each speaker. After that, the client 209b performs A / D conversion (not shown) on the voice signal and then performs a preprocessing process using the input signal.

Preprocessing process according to another preferred embodiment of the present invention is as described above.

The client 209b extracts the speaker component by preprocessing the A / D converted voice signal (step 1121). The member identifier, the word identifier and the speaker component are then sent to the network server system 201b (step 1123). Here, the client 209b may directly transmit the extracted speaker component to the speaker recognition server 203b.

According to another preferred embodiment of the present invention, the dispersion processing scheme is as described above.

The network server system 201b receives (step 1125) the words, word identifiers, and speaker components from the client. Thereafter, the member identifier, the word identifier, and the speaker component are transmitted (step 1125) to the speaker recognition server system 203b.

The speaker recognition server system 203b receives the member identifier, the word identifier, and the speaker component (step 1127). Of course, in another embodiment of the present invention, it may be received directly from the client (209b). Thereafter, reference pattern data is extracted through performing a learning process on the speaker components (step 1129).

According to another preferred embodiment of the present invention, the learning process is as described above.

The speaker recognition server system 203b stores the extracted reference pattern data in the voice information database 207b (step 1131). The voice information database 207b according to another preferred embodiment of the present invention is as described above.

Thereafter, the speaker recognition server system 203b determines whether there is a speaker component for which the learning process has not been performed (step 1133), and if the result of the determination is affirmative, extracts the reference pattern data (step 1129). If the result of the determination is negative, a completion notification is sent to the network server system 201b (step 1135). In another embodiment of the present invention, a completion notification may be sent directly to the client 209b.

The network server system 201b receives the completion notification and transmits it to the client 209b (step 1137).

The client 209b receives the completion notification from the network server system 201b and displays it on the display device (step 1139).

12 is a flowchart illustrating a speaker learning method according to another preferred embodiment of the present invention.

Referring to FIG. 12, the speaker learning method is generally the same as the speaker learning method as shown in FIG. Thus, a similar description may be substituted.

When the user transmits a connection request signal from the client 209b through an input device (step 1201), the network server system 201b receives the connection request signal (step 1203), and sends the message to the client 209b. The HTML document is sent (step 1205).

Herein, the eleventh HTML document may allow a user to input predetermined input items, for example, a member identifier, a name, a gender, an address, a social security number, an e-mail address, and the like into an HTML document for performing a general network service. Embedding an HTML document in the form of a form.

The client 209b receives the eleventh HTML document and displays it on the display device (step 1207). Thereafter, the user inputs the predetermined input items, for example, a member identifier, a name, a gender, an address, a social security number, an e-mail address, and the like through the input device and transmits the same to the network server system 201b. Step 1209).

The network server system 201b receives the predetermined input and stores it in the web site database 205b (step 1211). The web site database 205b according to another preferred embodiment of the present invention is as described above.

Thereafter, the network server system 201b determines a word group to learn from a plurality of word group tables constructed in advance using the predetermined input items, for example, gender, address, and e-mail address (step 1213). )do. Each of the plurality of pre-built word groups is composed of a plurality of words. The determination method is as described above.

The network server system 201b then transmits 1215 an twelfth HTML document to the client 209b. The twelfth HTML document embeds words belonging to a word group to be learned determined by the predetermined input item and corresponding pre-built word identifiers.

The client 209b receives the twelfth HTML document and displays it on the display device (step 1217). According to another preferred embodiment of the present invention, the client 209b receives a voice signal corresponding to each word from a user through a microphone. (Step 1219). It is desirable for the user to speak four or more times for each speaker. The client 209b then performs A / D conversion (not shown) on the voice signal to transmit a member identifier, a word identifier and the voice signal to the network server system 201b (step 1221). In another embodiment of the present invention, the client 209b may transmit the member identifier, the word identifier, and the voice signal to the speaker recognition server system 203b.

The network server system 201b receives the member identifier, the word identifier, and the voice signal from the client 209b and transmits it to the speaker recognition server system 203b (step 1223). The speaker recognition server system 203b receives the member identifier, the word identifier, and the voice signal (step 1225). Thereafter, the speaker recognition server system 203b extracts the speaker component through the preprocessing process (step 1227). Thereafter, the speaker component extracts reference pattern data by performing the learning process (step 1229).

The speaker recognition server system 203b stores the extracted reference pattern data in the voice information database 207b (step 1231). The voice information database 207b according to another preferred embodiment of the present invention is as described above.

Thereafter, the speaker recognition server system 203b determines whether there is a speaker component for which the learning process has not been performed (step 1233), and if the determination result is positive, the reference pattern data extraction step (step 1229). Returning, if the determination is negative, a completion notification is sent to the network server system 201b (step 1235). According to another embodiment of the present invention, the speaker recognition server system 203b sends the completion notification to the client. Send directly to 209b.

The network server system 201b sends the completion notification to the client 209b (step 1237).

The client 209b receives the completion notification from the network server system 201b and displays it on the display device (step 1239).

13 is a flowchart showing a speaker identification method according to another preferred embodiment of the present invention.

When the user transmits a connection request signal from the client 209b through an input device (step 1301), the network server system 201b receives the connection request signal (step 1303), and sends it to the client 209b (13th). The HTML document is sent (step 1305).

Herein, the thirteenth HTML document is an embedded HTML document in which a user can input a member identifier into an HTML document for performing a general network service.

The client 209b receives the thirteenth HTML document and displays it on a display device (step 1307). Thereafter, a member identifier is received from the user through the input device and transmitted to the network server system 201b (step 1309).

The network server system 201b receives the member identifier (step 1311) and retrieves the word group identifier stored in the web site database 205b corresponding to the member identifier (step 1313).

The word group is built in a dictionary, and the network server system 201b arbitrarily selects one word among a plurality of words belonging to the word group (step 1315). However, each time the speaker is identified, one of a plurality of words belonging to the word group may be selected in consideration of the time, the speaker's state, and the connection location. For example, when the user is in a nasal cold state, the user selects the remainder except for the non-negative form (for example, the word 'mom' includes the non-negative).

The network server system 201b then sends a fourteenth HTML document to the client 209b (step 1317). The fourteenth HTML document preprocesses a pre-built word identifier corresponding to the word and a voice signal in which the user pronounces the word in the client 209b to extract a speaker component corresponding to the word to the server. Embedding component software for transmission.

The client 209b receives the fourteenth HTML document and displays it on a display device (step 1319). According to another preferred embodiment of the present invention, the client 209b receives a voice signal corresponding to each word from a user through a microphone. (Step 1321). After that, the client 209b performs A / D conversion (not shown) on the voice signal and then performs a preprocessing process using the input signal.

Preprocessing process according to another preferred embodiment of the present invention is as described above.

The audio signal is input, and the speaker component is extracted (step 1323) through the preprocessing process.

Thereafter, the speaker component is transmitted to the network server system 201b (step 1325). In another embodiment of the present invention, the client 209b directly sends the speaker component to the speaker recognition server system 203b.

According to another preferred embodiment of the present invention, the dispersion processing scheme is as described above.

The network server system 201b receives the speaker component from the client 209b and transmits it (step 1327) to the speaker recognition server system 203b together with the member identifier and the word identifier.

Thereafter, the speaker recognition server system 203b receives the member identifier, the word identifier, and the speaker component (step 1329), and inputs the speaker component to perform the speaker verification process as shown in FIG. Speech pattern data is extracted (step 1330).

Then, the reference pattern data stored in the voice information database 207b corresponding to the word identifier and the voice pattern data is retrieved (step 1331).

Thereafter, the speaker recognition server system 203b determines whether the voice pattern data and the reference pattern data match (step 1333), and if the result of the determination is affirmative, notifies the network server system 201b of the permission notification. In step 1335, if the determination is negative, a rejection notification is sent to the network server system 201b (step 1337). According to another embodiment of the present invention, the speaker recognition server system 203b transmits the permission notification or the rejection notification to the client 209b.

The network server system 201b receives the grant notification and sends it to the client 209b (step 1339), or receives the reject notification and sends it to the client 209b (step 1341).

The client 209b receives the completion notification from the network server system 201b and displays it on the display device (step 1343) or receives the rejection notification and displays it on the display device (step 1345).

14 is a flowchart showing a speaker identification method according to another preferred embodiment of the present invention.

Referring to FIG. 14, the speaker learning method is generally the same as the speaker learning method as shown in FIG. Thus, a similar description may be substituted.

When the user transmits a connection request signal from the client 209b through the input device (step 1401), the network server system 201b receives the connection request signal (step 1403), and sends it to the client 209b. The HTML document is sent (step 1405).

Here, the fifteenth HTML document is an embedded HTML document in which a user can input a member identifier into an HTML document for performing a general network service.

The client 209b receives the fifteenth HTML document and displays it on the display device (step 1407). Thereafter, a member identifier is received from the user through the input device and transmitted to the network server system 201b (step 1409).

The network server system 201b receives the member identifier (step 1411) and retrieves the word group identifier stored in the web site database 205b corresponding to the member identifier (step 1413).

The word group is constructed in a dictionary, and the network server system 201b arbitrarily selects one word among a plurality of words belonging to the word group (step 1415). However, each time the speaker is identified, one of a plurality of words belonging to the word group may be selected in consideration of the time, the speaker's state, and the connection location. For example, when the user is in a nasal cold state, the user selects the remainder except for the non-negative form (for example, the word 'mom' includes the non-negative).

The network server system 201b then transmits a sixteenth HTML document to the client 209b (step 1417). The sixteenth HTML document is embedded with a pre-built word identifier corresponding to the word.

The client 209b receives the sixteenth HTML document and displays it on a display device (step 1419). According to another preferred embodiment of the present invention, the client 209b receives a voice signal corresponding to each word from a user through a microphone. (Step 1421). The client 209b then A / D converts the voice signal (not shown) and then transmits it to the network server system 201b (step 1423). According to another embodiment of the present invention, the client 209b directly transmits the voice signal to the speaker recognition server system 203b.

The network server system 201b receives the voice signal from the client and transmits the voice signal to the speaker recognition server system 203b together with the member identifier and the word identifier (step 1425).

The speaker recognition server system 203b receives the member identifier, the word identifier, and the voice signal (step 1427), extracts the speaker component corresponding to the word identifier through the preprocessing process, and then the speaker. The component is input, and the speech pattern data corresponding to the word identifier is extracted through the speaker confirmation process as shown in FIG. 9 (step 1429).

Thereafter, the speaker recognition server system 203b searches for reference pattern data stored in the speech information database 207b corresponding to the word identifier and the speech pattern data (step 1431).

Thereafter, the speaker recognition server system 203b determines whether or not the voice pattern data and the reference pattern data match (step 1433). If the result of the determination is affirmative, the speaker server server 203b notifies the network server system 201B of a permission notification. In step 1435, if the determination is negative, a rejection notification is sent to the network server system 201B (step 1437). According to another embodiment of the present invention, the speaker recognition server system 203b transmits the permission notification or the rejection notification to the client 209b.

The network server system 201b receives the permission notification and sends it to the client 209b (step 1439), or receives the reject notification and sends it to the client 209b (step 1441).

The client 209b receives the completion notification from the network server system 201b and displays it on the display device (step 1443) or receives the rejection notification and displays it on the display device (step 1445).

In accordance with a preferred embodiment of the present invention, a network-based speaker learning apparatus (not shown) is provided with means for receiving from the client 103 a predetermined input including at least a member identifier in a website database. . The apparatus determines one word group among a plurality of word groups pre-built in the web site database, and combines the pre-built word group identifier corresponding to the word group with the inputs to the web site database. Means for storing. Here, each word group is composed of a plurality of words.

In addition, the apparatus preprocesses a plurality of words belonging to the word group, a pre-established word identifier corresponding to each word, and voice signals in which the user pronounces the words in the client 103. Means for extracting the speaker components corresponding to the component component for transmission to the server 101 and embedding the component software in the second HTML document for transmission to the client 103. The preprocessing process is as described above.

And the apparatus comprises means for receiving word identifiers and speaker components corresponding to each word from the client 103, means for extracting the reference pattern data corresponding to each word by performing the learning process on the speaker components; Means for generating a reference pattern identifier corresponding to each of the reference pattern data and storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database.

According to another preferred embodiment of the present invention, the network-based speaker learning apparatus has means for receiving from the client 103 a predetermined input including at least a member identifier and storing it in a website database, wherein the website database is provided. Means for determining one word group from among a plurality of word groups previously built in, and storing in said web site database a dictionary word group identifier corresponding to said word group in combination with said input. Here, each word group is composed of a plurality of words.

The apparatus further includes means for embedding a plurality of words belonging to the word group and a pre-built word identifier corresponding to each word into a fourth HTML document and transmitting the same to the client 103. Means for receiving word identifiers and speech signals corresponding to each word from 103, and means for preprocessing the speech signals to extract speaker components corresponding to each word. The preprocessing process is as described above.

The apparatus may further include means for extracting reference pattern data corresponding to each word through performing a learning process on the speaker components, generating a reference pattern identifier corresponding to each reference pattern data, and then generating the member identifier, Means for storing the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database 207b.

According to another preferred embodiment of the present invention, a network-based speaker identification apparatus (not shown) includes means for receiving a member identifier from the client 103 and a word group identifier stored in a web site database corresponding to the member identifier. Means for searching, selecting a word of a plurality of words belonging to said word group.

The device is a component for preprocessing the word, a word identifier corresponding to the word, and a voice signal in which the user pronounces the word to extract voice pattern data corresponding to the word and transmit the same to the server 101. Means for embedding component software into a sixth HTML document and transmitting it to the client 103, means for receiving voice pattern data corresponding to the word from the client 103, and voice corresponding to the word identifier Means for retrieving reference pattern data stored in the information database 207b, means for determining whether the voice pattern data and the reference pattern data match, and if the result of the determination is affirmative, sending a permission notification to the client 103. Means and means for sending a rejection notification to the client 103 if the determination result is negative. The.

According to another preferred embodiment of the present invention, a network-based speaker identification apparatus (not shown) includes means for receiving a member identifier from the client 103 and a word group identifier stored in a web site database corresponding to the member identifier. Means for searching, means for selecting one word among a plurality of words belonging to the word group, embedding the word and a word identifier corresponding to the word in an eighth HTML document and transmitting it to the client 103. Means, means for receiving a speech signal corresponding to the word from a client 103, means for preprocessing the speech signal to extract speech pattern data corresponding to the word, speech information corresponding to the word identifier Means for retrieving reference pattern data stored in a database 207b, the speech pattern data and the reference Means for determining whether the turn data matches, means for sending a grant notification to the client 103 if the determination is affirmative, and means for sending a reject notification to the client 103 if the determination is negative. .

According to another preferred embodiment of the present invention, a network-based speaker learning apparatus (not shown) in the network server system 201b receives a predetermined input from the client 209b including at least a member identifier from the client 209b. Means for storing in a site database, a word group of a plurality of word groups pre-built in the web site database, and combining the input word with a pre-built word group identifier corresponding to the word group Means for storing in a database. Here, each word group is composed of a plurality of words.

Further, the apparatus preprocesses a plurality of words belonging to the word group, a pre-built word identifier corresponding to each word, and a voice signal in which the user pronounces the words in the client 209b. Means for embedding component software in the tenth HTML document for transmission and extracting the corresponding speaker components into the tenth HTML document and transmitting the same to the client 209b, a member identifier from the client 209b, and the respective components. Means for receiving a word identifier corresponding to a word and a speaker component corresponding to each word and transmitting them to the speaker recognition server system 203b and receiving a learning completion notification from the speaker recognition server system 203b to receive the client ( Means for transmitting to 209b).

According to another preferred embodiment of the present invention, a network-based speaker learning apparatus (not shown) in a speaker recognition system may be a member identifier inputted directly from a client 209b or via a network server system 201b, respectively. Means for receiving word identifiers and speaker components corresponding to the word of, means for extracting the reference component data corresponding to each word by performing the learning process on the speaker components and a reference corresponding to the respective reference pattern data Means for storing the member identifier, the reference pattern data, the reference pattern identifier, and the word identifier in the voice information database 207b after generating the pattern identifier.

According to another preferred embodiment of the present invention, the network-based speaker learning apparatus (not shown) in the network server system 201b receives a predetermined input from the client 209b including at least a member identifier from the web 209b. Means for storing in a site database, a word group of a plurality of word groups pre-built in the web site database, and combining the input word with a pre-built word group identifier corresponding to the word group Means for storing in a database. Here, each word group is composed of a plurality of words.

The apparatus further comprises means for embedding a plurality of words belonging to the word group and a pre-built word identifier corresponding to each word into a twelfth HTML document and transmitting it to the client 209b, from the client 209b. Means for receiving a member identifier, a word identifier corresponding to each word, and a speech signal corresponding to each word and transmitting them to the speaker recognition server system 203b and notification of learning completion from the speaker recognition server system 203b. Means for receiving and transmitting to the client 209b.

According to another preferred embodiment of the present invention, a network-based speaker learning apparatus (not shown) in a speaker recognition system may be a member identifier inputted directly from a client 209b or via a network server system 201b, respectively. Means for receiving word identifiers and speech signals corresponding to the word of, means for preprocessing the speech signals to extract speaker components corresponding to each word, and performing the learning process on the speaker components, respectively. Means for extracting reference pattern data corresponding to a word, and generating a reference pattern identifier corresponding to each reference pattern data, and then converting the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier into a voice information database. Means for storing in 207b).

According to another preferred embodiment of the present invention, a network-based speaker identification apparatus (not shown) in a network server system 201b is provided with means for receiving a member identifier from a client 209b, a web corresponding to the member identifier. Means for retrieving a word family identifier stored in a site database, means for selecting one word among a plurality of words belonging to the word family, the word, a word identifier corresponding to the word, and the user at the client 209b. Preprocessing a speech signal having a pronounced sound, and extracting the speech pattern data corresponding to the word, and embedding component software for transmission to the network server system 201b in a fourteenth HTML document. Means for transmitting to the client 209b, and voice pattern data from the client 209b. Means for receiving and transmitting the speech pattern data, the member identifier and the word identifier to the speaker recognition server system 203b, receiving a permission notification from the speaker recognition server system 203b and transmitting it to the client 209b. Means for receiving a rejection notification from the speaker recognition server system 203b and transmitting it to the client 209b.

According to another preferred embodiment of the present invention, a network-based speaker identification apparatus (not shown) in a speaker recognition system may include a member identifier, a word input directly from a client 209b or via a network server system 201b. Means for receiving an identifier and speech pattern data, means for retrieving reference pattern data stored in a speech information database 207b corresponding to the word identifier, and means for determining whether the speech pattern data and the reference pattern data match. Means for sending an authorization notification to the client 209b or the network server system 201b if the determination result is affirmative, and a rejection notification to the client 209b or the network server system 201b if the determination result is negative. Means for transmitting.

According to another preferred embodiment of the present invention, a network-based speaker identification apparatus (not shown) in a network server system 201b is provided with means for receiving a member identifier from a client 209b, a web corresponding to the member identifier. Means for retrieving a word group identifier stored in a site database, means for selecting one word among a plurality of words belonging to the word group, and embedding the word and a word identifier corresponding to the word in a sixteenth HTML document. Means for receiving the voice signal from the client 209b, transmitting the voice signal, the member identifier and the word identifier to the speaker recognition server system 203b, and the speaker recognition. Means for receiving an authorization notice from server system 203b and sending it to the client 209b; and Party receiving the rejection notice from the recognition server system (203b) includes means for transmitting to said client (209b).

According to another preferred embodiment of the present invention, a network-based speaker identification apparatus (not shown) in a speaker recognition system may include a member identifier, a word input directly from a client 209b or via a network server system 201b. Means for receiving an identifier and a speech signal, means for preprocessing the speech signal to extract speech pattern data corresponding to the word, and searching for reference pattern data stored in the speech information database 207b corresponding to the word identifier. Means for determining whether or not the voice pattern data and the reference pattern data match, and if the determination result is affirmative, means for transmitting a permission notification to the client 209b or the network server system 201b, and If the determination is negative, the client 209b or the network server system. And a means for sending a rejection notice to (201b).

In the case of electronic commerce payment, if the identification is confirmed by the speaker according to the present invention, the identification information according to the present invention stores the voice information corresponding to the pronunciation component in the voice information data. In case of filing, the user can restore the pronunciation component to the user and remove the factor preventing the activation of the transaction. Here, the pronunciation component corresponds to each order when the user makes a plurality of orders.

The network-based speaker learning method and speaker identification method according to the present invention can be realized by a system including a web site database, a voice signal database and a memory in which a program is stored, and a processor coupled to the memory to execute the program. have. Here, the processor may execute a method including the respective steps by the program.

The present invention is not limited to the above embodiments, and many variations are possible by those skilled in the art within the spirit of the present invention.

As described above, the present invention relates to a network-based speaker learning and speaker identification method and apparatus, and in particular, a letter-directed speaker learning and speaker identification method using a word group composed of words representing personal characteristics of the speaker well. And an apparatus.

According to the present invention, without using a separate password to log in to a site providing information using the Internet, each individual without having to install additional equipment other than the basic computer environment among the unique characteristics each individual has A speaker learning and speaker identification method and apparatus using the voice characteristic of the present invention can be provided.

According to the present invention, a method and apparatus for character-directed speaker learning and speaker identification using a word group composed of words well representing the speaker's personal characteristics according to the speaker's gender and age, in order to improve speaker recognition and impersonate defense due to recording. Can be provided.

In addition, the present invention provides a method and apparatus for distributed speaker design and speaker identification designed to prevent the load from being concentrated even when many users are connected to the system because the number of simultaneous users cannot be predicted on the Internet. can do.

And, according to the present invention, learn a plurality of words, speaker learning and speaker identification that can be confirmed by the speaker by selecting one word from the words learned in consideration of the time, speaker status, connection location, etc. each time the speaker is confirmed According to the present invention, when a contract is established by a customer placing a certain order in an electronic commerce, the voice signal corresponding to the order is received, processed and stored therein, When the customer knows whether the contract is established or ordered, the voice signal may be restored to the customer and provided to the customer, thereby providing a voice restoration method capable of enhancing the stability of the author commerce.

Claims (93)

In a network-based user authentication method for authenticating a user connected using a microphone at a server, Registering the user by storing user identification information together with user voice characteristic information obtained by analyzing the voice signal of the user; Receiving a user authentication request from an authentication requester directly or indirectly through the network; Selecting and transmitting a sentence or sentence corresponding to the user's state from a pre-built word group or sentence group to the authentication requester; Receiving a voice signal of the authentication requestor for the word or sentence or voice characteristic information of the authentication requester extracted from the voice signal through a network; And Determining whether the voice signal or voice property information of an authentication requester extracted from the voice signal is input by the user using the registered user voice property information. User authentication method using a network-based word group or sentence group in the server comprising a. In the network-based speaker learning method for authenticating a user connected using a microphone in a server, Receiving from the client predetermined input including at least a member identifier and storing it in a website database; Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group Storing the data in the web site database in combination with the details; Preprocessing a plurality of words belonging to the word group, a pre-established word identifier corresponding to each word, and voice signals in which a user corresponding to the client pronounces the words in a client to correspond to the respective words Extracting speaker components to be transmitted to the client by embedding component software in a first document for transmission to a server; Receiving word identifiers and speaker components corresponding to each word from the client; Extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; And Generating a reference pattern identifier corresponding to each of the reference pattern data, and then storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker learning method using network-based word group for user authentication. The method of claim 3, Speaker learning method using a network-based word group for user authentication in the server Receiving a connection request signal from the client, and transmitting a second document in a form of a form for inputting a predetermined input item to the user to the client; Receiving the predetermined input from the client and storing it in the web site database; Determining whether there is a speaker component for which a learning process is not performed; If there is a speaker component for which a learning process is not performed as a result of the determination, shifting to extracting the reference pattern data; And If there is no speaker component for which the learning process has not been performed as a result of the determination, a speaker learning method using a network-based word group for user authentication in a server, further comprising transmitting a completion notification to the client. . The method of claim 3, The predetermined input items are a member identifier, a name, a gender, an address, a social security number, an e-mail address, and the speaker learning method using a network-based word group for user authentication in a server. The method of claim 5, The determining of the one word group corresponds to at least one of a gender, an address, and an e-mail address. The speaker learning method using a network-based word group for user authentication in a server. It can be executed by the digital processing apparatus to perform the speaker learning method using the network-based word group in the client corresponding to performing the network-based speaker learning method for authenticating the connected user using the microphone in the server. A recording medium in which a program of instructions is tangibly embodied, and which can be read by a digital processing apparatus, Speaker learning method using a network-based word group in the client, Preprocessing words belonging to a specific word group from the server, a pre-built word identifier corresponding to each word, and a voice signal in which the user corresponding to the client pronounces the words. Receiving a first document embedding component software for extracting speaker components corresponding to a word and transmitting the same to a server; Displaying the first document on a display device; Receiving a voice signal corresponding to each word from the user; Extracting a speaker component from the speech signal corresponding to each word; And And sending the word identifier and speaker components corresponding to each word to the server. The method of claim 7, wherein Speaker learning method using the network-based word group in the client Transmitting a connection request signal input by the user by an input means to the server; Receiving a second document in the form of a form from which the user can input a predetermined input item from the server and displaying the second document on the display device; Receiving the predetermined input item from the user and transmitting the predetermined input item to the server; And And receiving the speaker learning completion notification from the server and displaying the speaker learning completion notice on the display device. The method of claim 7, wherein And the predetermined input item is a member identifier, a name, a gender, an address, a social security number, and an e-mail address. The method of claim 7, wherein Speaker learning method using the network-based word group in the client And converting the voice signal into a digital signal. 10. The recording medium of claim 1, further comprising: converting the voice signal to a digital signal. In the network-based speaker learning method for authenticating a user connected using a microphone in a server, Receiving from the client predetermined input including at least a member identifier and storing it in a website database; Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group Storing the data in the web site database in combination with the details; Embedding a plurality of words belonging to the word group and a pre-built word identifier corresponding to each word into a first document and transmitting the same to the client; Receiving word identifiers and voice signals corresponding to the respective words from the client; Preprocessing the speech signals to extract speaker components corresponding to each word; Extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; And Generating a reference pattern identifier corresponding to each of the reference pattern data, and then storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker learning method using network-based word group for user authentication. The method of claim 11, Speaker learning method using a network-based word group for user authentication in the server Receiving a connection request signal from the client, and transmitting a second document in a form of a form for inputting a predetermined input item to the user to the client; Receiving the predetermined input from the client and storing it in the web site database; Determining whether there is a speaker component for which a learning process is not performed; If there is a speaker component for which a learning process is not performed as a result of the determination, shifting to extracting the reference pattern data; And If there is no speaker component for which the learning process has not been performed as a result of the determination, a speaker learning method using a network-based word group for user authentication in a server, further comprising transmitting a completion notification to the client. . The method of claim 11, The predetermined input items are a member identifier, a name, a gender, an address, a social security number, and an e-mail address. The method of claim 13, Determining the one word group is determined in correspondence with at least one of gender, address, and e-mail address. Speaker learning method using a network-based word group for user authentication in the server. It can be executed by the digital processing apparatus to perform the speaker learning method using the network-based word group in the client corresponding to performing the network-based speaker learning method for authenticating the connected user using the microphone at the server. A recording medium in which a program of instructions is tangibly embodied, and which can be read by a digital processing apparatus, Speaker learning method using a network-based word group in the client, Receiving, from the server, a first document in which words belonging to a specific word group and a pre-built word identifier corresponding to each word are embedded; Displaying the first document on a display device; Receiving a voice signal corresponding to each word from a user; And And transmitting word identifiers and voice signals corresponding to the respective words to the server. The method of claim 15, Speaker learning method using the network-based word group in the client Transmitting a connection request signal input by the user by an input means to the server; Receiving a second document in the form of a form from which the user can input a predetermined input item from the server and displaying the second document on the display device; Receiving the predetermined input item from the user and transmitting the predetermined input item to the server; And And receiving the speaker learning completion notification from the server and displaying the speaker learning completion notice on the display device. The method of claim 15, And the predetermined input item is a member identifier, a name, a gender, an address, a social security number, and an e-mail address. The method of claim 15, Speaker learning method using the network-based word group in the client And converting the voice signal into a digital signal. 10. The recording medium of claim 1, further comprising: converting the voice signal to a digital signal. In the network-based speaker identification method for authenticating a user connected using a microphone in the server, Receiving a member identifier from a client; Retrieving a word group identifier previously stored in a web site database corresponding to the member identifier; Selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier; A first component software for preprocessing the word, the word identifier corresponding to the word, and a voice signal in which the user pronounces the word to extract a voice signal corresponding to the word and transmit it to a server Embedding in a document and sending it to the client; Receiving a voice signal corresponding to the word from the client; Extracting voice pattern data corresponding to the word identifier by inputting the voice signal; Retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Determining whether the voice pattern data and the reference pattern data match; Sending a permission notification to the client if the voice pattern data and the reference pattern data match as a result of the determination; And And if the voice pattern data and the reference pattern data do not match, transmitting a rejection notification to the client, wherein the speaker check method using the network-based word group for user authentication in the server. . The method of claim 19, Speaker verification method using a network-based word group for user authentication in the server Receiving a connection request signal from the client, and transmitting to the client a second document of the form (Form) form for inputting the member identifier to the user, characterized in that for the user authentication in the server Speaker identification method using network based word group. The method of claim 19, Speaker selection method using a network-based word group for user authentication in the server, characterized in that the selection of the single word random (random). The method of claim 19, The selecting of the single word may include selecting one of a plurality of words belonging to the word group in consideration of the time at the speaker confirmation, the speaker's state, and the connection location. being- Speaker verification method using network-based word group for user authentication on server. It can be executed by the digital processing apparatus to perform the speaker identification method using the network-based word group in the client corresponding to performing the network-based speaker identification method for authenticating the connected user using the microphone at the server. A recording medium in which a program of instructions is tangibly embodied, and which can be read by a digital processing apparatus, Speaker verification method using a network-based word group in the client, Transmitting a connection request signal input by a user by an input means to the server; Receiving, from the server, a second document in the form of a form capable of inputting a member identifier from the server to the display device; Transmitting the member identifier input by the user by an input means to the server; Preprocessing a word corresponding to the member identifier, a word identifier corresponding to the word, and a voice signal in which the user pronounces the word from the server to extract a speaker component corresponding to the word and transmit the same to the server Receiving a first document in which component software is embedded; Displaying the first document on the display device; Receiving a voice signal corresponding to the word from the user; Extracting a speaker component from a speech signal corresponding to the word; Sending the speaker component to the server; If the permission notification is received from the server, displaying the permission notification on the display device; And And displaying the rejection notice on the display device when receiving the rejection notice from the server. The method of claim 23, wherein Speaker verification method using the network-based word group in the client And converting the voice signal into a digital signal. 10. The recording medium of claim 1, further comprising: converting the voice signal to a digital signal. In the network-based speaker identification method for authenticating a user connected using a microphone in the server, Receiving a member identifier from a client; Retrieving a word group identifier stored in a web site database corresponding to the member identifier; Selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier; Embedding the word and a word identifier corresponding to the word in a first document and transmitting it to the client; Receiving a voice signal corresponding to the word from the client; Preprocessing the voice signal to extract a speaker component corresponding to the word; Extracting voice pattern data corresponding to the word by using the speaker component as an input; Retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Determining whether the voice pattern data and the reference pattern data match; Sending a permission notification to the client if the voice pattern data and the reference pattern data match as a result of the determination; And And if the voice pattern data and the reference pattern data do not match, transmitting a rejection notification to the client, wherein the speaker check method using the network-based word group for user authentication in the server. . The method of claim 25, Speaker verification method using a network-based word group for user authentication in the server Receiving a connection request signal from the client, and sending a second document in the form (Form) form for inputting a member identifier to the user to the client, characterized in that the user authentication at the server Speaker identification method using network-based word group. The method of claim 25, Speaker selection method using a network-based word group for user authentication in the server, characterized in that the selection of the single word random (random). The method of claim 25, The selecting of the single word may include selecting one of a plurality of words belonging to the word group in consideration of the time at the speaker confirmation, the speaker's state, and the connection location. being- Speaker verification method using network-based word group for user authentication on server. It can be executed by the digital processing apparatus to perform the speaker identification method using the network-based word group in the client corresponding to performing the network-based speaker identification method for authenticating the connected user using the microphone at the server. A recording medium in which a program of instructions is tangibly embodied, and which can be read by a digital processing apparatus, The network-based speaker identification method using the word group in the client, Transmitting a connection request signal input by a user by an input means to the server; Receiving, from the server, a second document in the form of a form capable of inputting a member identifier from the server to the display device; Transmitting the member identifier input by the user by the input means to the server; Receiving a first document in which the word corresponding to the member identifier and the word identifier corresponding to the word are embedded from the server; Displaying the first document on a display device; Receiving a voice signal corresponding to the word from the user; Transmitting the voice signal to the server; If the permission notification is received from the server, displaying the permission notification on the display device; And And displaying the rejection notice on the display device when receiving the rejection notice from the server. The method of claim 29, Speaker verification method using the network-based word group in the client And converting the voice signal into a digital signal. 10. The recording medium of claim 1, further comprising: converting the voice signal to a digital signal. In the network-based speaker learning method for authenticating a user connected by using a microphone of a network server system among server systems, Receiving from the client predetermined input including at least a member identifier and storing it in a website database; Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group Storing the data in the web site database in combination with the details; Pre-processing a plurality of words belonging to the word group, a pre-built word identifier corresponding to each word, and a voice signal in which the user pronounces the words at the client to extract speaker components corresponding to each word Embedding component software for transmission to the server into a first document and transmitting it to the client; Receiving a member identifier, a word identifier corresponding to each word, and a speaker component corresponding to each word from the client and transmitting them to a speaker recognition server system; And And receiving a learning completion notification from the speaker recognition server system and transmitting the notification to the client. The speaker learning method using a network-based word group for user authentication in a network server system among the server systems. The method of claim 31, wherein Speaker learning method using a network-based word group for user authentication in a network server system of the server system Receiving a connection request signal from the client and transmitting a second document in a form of a form capable of inputting a predetermined input item to the user to the client; And Receiving the predetermined input from the client and storing the predetermined input item in the web site database; a speaker learning method using a network-based word group for user authentication in a network server system among server systems . The method of claim 31, wherein The predetermined input items are a member identifier, a name, a gender, an address, a social security number, and an e-mail address. The speaker learning method using a network-based word group for user authentication in a network server system among server systems. The method of claim 33, wherein Determining the single word group corresponds to at least one of a gender, an address, and an e-mail address. A speaker learning method using a network-based word group for user authentication in a network server system among server systems . In the network-based speaker learning method for authenticating a connected user using a microphone in a speaker recognition system among server systems, Receiving a member identifier, word identifiers corresponding to each word, and speaker components entered directly from the client or via a network server system; Extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; And Generating a reference pattern identifier corresponding to each of the reference pattern data, and then storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker learning method using network-based word group for user authentication in Chinese speaker recognition server system. 36. The method of claim 35 wherein Speaker learning method using a network-based word group for user authentication in a speaker recognition server system of the server system Determining whether there is a speaker component for which a learning process is not performed; If there is a speaker component for which a learning process is not performed as a result of the determination, shifting to extracting the reference pattern data; And If there is no speaker component for which the learning process has not been performed as a result of the determination, transmitting a completion notification to the client or the network server system. Speaker learning method using network-based word group for authentication. The speaker learning method using a network-based word group in a client is equivalent to performing a network-based speaker learning method for authenticating a connected user using a microphone in a network server system and a speaker recognition server system among server systems. A program of instructions that can be executed by a digital processing device to perform is tangibly embodied, and in a recording medium that can be read by a digital processing device, Speaker learning method using a network-based word group in the client, Pre-processing words belonging to a specific word group from the network server system, a pre-built word identifier corresponding to each word, and a voice signal in which the user pronounces the words in the client corresponding to each word Receiving a first document embedding component software for extracting speaker components and transmitting the same to the server; Displaying the first document on a display device; Receiving a voice signal corresponding to each word from the user; Extracting a speaker component from the speech signal corresponding to each word; And Transmitting the member identifier, word identifiers corresponding to each word, and speaker components to the network server system or the speaker recognition server system. The method of claim 37, Speaker learning method using the network-based word group in the client Transmitting a connection request signal input by the user by an input means to the network server system; Receiving a second document in the form of a form from which the user can input predetermined input from the network server system and displaying the second document on the display device; Receiving the predetermined input from the user and transmitting the predetermined input to the network server system; And And receiving the speaker learning completion notification from the network server system or the speaker recognition server system and displaying the speaker learning completion notification on the display device. The method of claim 37, And the predetermined input item is a member identifier, a name, a gender, an address, a social security number, and an e-mail address. The method of claim 37, Speaker learning method using the network-based word group in the client And converting the voice signal into a digital signal. 10. The recording medium of claim 1, further comprising: converting the voice signal to a digital signal. In the network-based speaker learning method for authenticating a user connected by using a microphone of a network server system among server systems, Receiving from the client predetermined input including at least a member identifier and storing it in a website database; Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group Storing the data in the web site database in combination with the details; Embedding a plurality of words belonging to the word group and a pre-built word identifier corresponding to each word into a first document and transmitting the same to the client; Receiving a member identifier, a word identifier corresponding to each word, and a voice signal corresponding to each word from the client and transmitting the same to the speaker recognition server system; And And receiving a learning completion notification from the speaker recognition server system and transmitting the notification to the client. The speaker learning method using a network-based word group for user authentication in a network server system among the server systems. The method of claim 41, wherein Speaker learning method using a network-based word group for user authentication in a network server system of the server system Receiving a connection request signal from the client and transmitting a second document in a form of a form capable of inputting a predetermined input item to the user to the client; And Receiving the predetermined input from the client and storing the predetermined input item in the web site database; a speaker learning method using a network-based word group for user authentication in a network server system among server systems . The method of claim 41, wherein The predetermined input items are a member identifier, a name, a gender, an address, a social security number, and an e-mail address. The speaker learning method using a network-based word group for user authentication in a network server system among server systems. The method of claim 43, Determining the single word group corresponds to at least one of a gender, an address, and an e-mail address. A speaker learning method using a network-based word group for user authentication in a network server system among server systems . In the network-based speaker learning method for authenticating a connected user using a microphone in a speaker recognition system among server systems, Receiving a member identifier, word identifiers corresponding to each word, and voice signals input directly from the client or via a network server system; Preprocessing the voice signals to extract speaker components corresponding to each word; Extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; And Generating a reference pattern identifier corresponding to each of the reference pattern data, and then storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker learning method using network-based word group for user authentication in Chinese speaker recognition server system. The method of claim 45, Speaker learning method using a network-based word group for user authentication in a speaker recognition server system of the server system Determining whether there is a speaker component for which a learning process is not performed; If there is a speaker component for which a learning process is not performed as a result of the determination, shifting to extracting the reference pattern data; And If there is no speaker component for which the learning process has not been performed as a result of the determination, transmitting a completion notification to the client or the network server system. Speaker learning method using network-based word group for authentication. The speaker learning method using a network-based word group in a client is equivalent to performing a network-based speaker learning method for authenticating a connected user using a microphone in a network server system and a speaker recognition server system among server systems. A program of instructions that can be executed by a digital processing device to perform is tangibly embodied, and in a recording medium that can be read by a digital processing device, Speaker learning method using a network-based word group in the client, Receiving, from the network server system, a first document in which words belonging to a specific word group and a pre-built word identifier corresponding to each word are embedded; Displaying the first document on a display device; Receiving a voice signal corresponding to each word from a user; And Transmitting the member identifier, word identifiers corresponding to each word, and voice signals to the network server system or the speaker recognition server system. The method of claim 47, Speaker learning method using the network-based word group in the client Transmitting a connection request signal input by the user by an input means to the network server system; Receiving a second document in the form of a form from which the user can input predetermined input from the network server system and displaying the second document on the display device; Receiving the predetermined input from the user and transmitting the predetermined input to the network server system; And And receiving the speaker learning completion notification from the network server system or the speaker recognition server system and displaying the speaker learning completion notification on the display device. The method of claim 47, And the predetermined input item is a member identifier, a name, a gender, an address, a social security number, and an e-mail address. The method of claim 47, Speaker learning method using the network-based word group in the client And converting the voice signal into a digital signal. 10. The recording medium of claim 1, further comprising: converting the voice signal to a digital signal. In the network-based speaker identification method for authenticating a user connected by using the microphone of the network server system of the server system, Receiving a member identifier from a client; Retrieving a word group identifier stored in a web site database corresponding to the member identifier; Selecting one word among a plurality of words belonging to the word group; A component software for preprocessing the word, a word identifier corresponding to the word, and a voice signal in which the user pronounces the word in the client, extracting a speaker component corresponding to the word and transmitting it to the network server system ( embedding component software into a first document and transmitting the same to the client; Receiving a speaker component from the client, and transmitting the speaker component, the member identifier, and the word identifier to a speaker recognition server system; Receiving a permission notification from the speaker recognition server system and transmitting it to the client; And And receiving the rejection notification from the speaker recognition server system and transmitting the same to the client. The method of claim 51, Speaker verification method using a network-based word group for user authentication in a network server system Receiving a connection request signal from the client, the network server system of the server system, characterized in that it further comprises the step of transmitting to the client a second document in the form (Form) form Speaker Verification Method Using Network based Word Group for User Authentication in. The method of claim 51, The method of claim 1, wherein selecting one word is randomly selected. The method of claim 51, The selecting of the single word may include selecting one of a plurality of words belonging to the word group in consideration of the time at the speaker confirmation, the speaker's state, and the connection location. being- Speaker verification method using network based word group for user authentication in network server system among server systems. In the network-based speaker identification method for authenticating a connected user using a microphone in a speaker recognition system of the server system, Receiving a member identifier, a word identifier and a speaker component entered directly from a client or via a network server system; Extracting speech pattern data corresponding to the word identifier by inputting the speaker component; Retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Determining whether the voice pattern data and the reference pattern data match; Transmitting a permission notification to the client or the network server system when the voice pattern data and the reference pattern data match as a result of the determination; And And if the voice pattern data and the reference pattern data do not match, transmitting a rejection notification to the client or the network server system. Speaker verification method using network-based word group. A speaker identification method using a network-based word group in a client corresponding to performing a network-based speaker identification method for authenticating a connected user using a microphone in a network server system and a speaker recognition server system among server systems is described. A program of instructions that can be executed by a digital processing device to perform is tangibly embodied, and in a recording medium that can be read by a digital processing device, Speaker verification method using a network-based word group in the client, Transmitting a connection request signal input by a user by an input means to the network server system; Receiving a second document in a form of a form capable of inputting a member identifier from the network server system and displaying the second document on a display device; Transmitting the member identifier input by the user by the input means to the network server system; Preprocessing the word corresponding to the member identifier, the word identifier corresponding to the word, and the voice signal of the user pronouncing the word from the network server system to extract the speaker component corresponding to the word and the network server Receiving a first document embedding component software for transmission to a system or said speaker recognition server system; Displaying the first document on the display device; Receiving a voice signal corresponding to the word from the user; Extracting a speaker component from a speech signal corresponding to the word; Sending the speaker component to the network server system or the speaker recognition server system; Displaying the permission notification on the display device when receiving the permission notification from the network server system or the speaker recognition server system; And And displaying the rejection notification on the display device when receiving the rejection notification from the network server system or the speaker recognition server system. The method of claim 56, wherein Speaker verification method using the network-based word group in the client And converting the voice signal into a digital signal. 10. The recording medium of claim 1, further comprising: converting the voice signal to a digital signal. In the network-based speaker identification method for authenticating a user connected by using the microphone of the network server system of the server system, Receiving a member identifier from a client; Retrieving a word group identifier stored in a web site database corresponding to the member identifier; Selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier; Embedding the word and a word identifier corresponding to the word in a first document and transmitting it to the client; Receiving a voice signal from the client and transmitting the voice signal, the member identifier and the word identifier to a speaker recognition server system; Receiving a permission notification from the speaker recognition server system and transmitting it to the client; And And receiving the rejection notification from the speaker recognition server system and transmitting the same to the client. The method of claim 58, Speaker verification method using a network-based word group for user authentication in a network server system Receiving a connection request signal from the client, the network server system of the server system, characterized in that it further comprises the step of transmitting to the client a second document in the form (Form) form Speaker Verification Method Using Network based Word Group for User Authentication in. The method of claim 58, The method of claim 1, wherein selecting one word is randomly selected. The method of claim 58, The selecting of one word may include selecting one of a plurality of words belonging to the word group in consideration of a time of speaker identification, a speaker state, and a connection location. Speaker verification method using network-based word group for user authentication. In the network-based speaker identification method for authenticating a connected user using a microphone in a speaker recognition system of the server system, Receiving a member identifier, a word identifier and a voice signal input directly from the client or via a network server system; Preprocessing the voice signal to extract a speaker component corresponding to the word; Extracting voice pattern data corresponding to the word by using the speaker component as an input; Retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Determining whether the voice pattern data and the reference pattern data match; Transmitting a permission notification to the client or the network server system if the voice pattern data and the reference pattern data match as a result of the determination; And And if the voice pattern data and the reference pattern data do not match, sending a rejection notification to the client or the network server system. Speaker identification method using network-based word group. A speaker identification method using a network-based word group in a client corresponding to performing a network-based speaker identification method for authenticating a connected user using a microphone in a network server system and a speaker recognition server system among server systems is described. A program of instructions that can be executed by a digital processing device to perform is tangibly embodied, and in a recording medium that can be read by a digital processing device, Speaker verification method using a network-based word group in the client, Transmitting a connection request signal input by a user by an input means to the network server system; Receiving a second document in a form of a form capable of inputting a member identifier from the network server system and displaying the second document on a display device; Transmitting the member identifier input by the user by the input means to the network server system; Receiving a first document in which the word corresponding to the member identifier and the word identifier corresponding to the word are embedded from the network server system; Displaying the first document on the display device; Receiving a voice signal corresponding to the word from the user; Transmitting the voice signal to the network server system or the speaker recognition server system; Displaying the permission notification on the display device when receiving the permission notification from the network server system or the speaker recognition server system; And And displaying the rejection notification on the display device when receiving the rejection notification from the network server system or the speaker recognition server system. The method of claim 63, wherein Speaker verification method using the network-based word group in the client And converting the voice signal into a digital signal. 10. The recording medium of claim 1, further comprising: converting the voice signal to a digital signal. In the network-based user authentication device for authenticating a user connected using a microphone, Means for registering the user by storing the user identification information together with the user voice characteristic information obtained by analyzing the user's voice signal; Means for receiving a user authentication request from an authentication requester directly or indirectly through the network; Selecting a word or sentence corresponding to the user's state from a pre-built word group or sentence group and transmitting the selected word or sentence to the authentication requester; Means for receiving a voice signal of the authentication requestor for the word or sentence or voice characteristic information of the authentication requester extracted from the voice signal through the network; Means for determining whether the voice signal or voice property information of an authentication requester extracted from the voice signal is input by the user using the registered user voice property information. User authentication device using a network-based word group or sentence group having a. 66. The method of claim 65, And the word is selected based on personal information of the user registered in advance corresponding to the authentication requestor. In the network-based speaker learning apparatus for authenticating a user connected using a microphone, Means for receiving from the client predetermined input including at least a member identifier and storing it in a website database; Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group Means for storing in the web site database in conjunction with the details; Preprocessing a plurality of words belonging to the word group, a pre-established word identifier corresponding to each word, and voice signals in which the user corresponding to the client pronounces the words in the client corresponding to each word Means for extracting speaker components and embedding component software in a first document for transmission to a server and transmitting the same to the client; Means for receiving word identifiers and speaker components corresponding to each word from the client; Means for extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; And And means for storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database after generating a reference pattern identifier corresponding to each of the reference pattern data. Speaker learning device using network-based word group. In the network-based speaker learning apparatus for authenticating a user connected using a microphone, Means for receiving from the client predetermined input including at least a member identifier and storing it in a website database; Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group Means for storing in the web site database in conjunction with the details; Means for embedding a plurality of words belonging to the word group and a pre-built word identifier corresponding to each word into a first document and transmitting the same to the client; Means for receiving word identifiers and voice signals corresponding to each word from the client; Means for preprocessing the speech signals to extract speaker components corresponding to each word; Means for extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; And And means for storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database after generating a reference pattern identifier corresponding to each of the reference pattern data. Speaker learning device using network-based word group. In the network-based speaker identification device for authenticating a user connected using a microphone, Means for receiving a member identifier from a client; Means for retrieving a word-group identifier stored in a web site database corresponding to the member identifier; Selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier; A first component software for preprocessing the word, the word identifier corresponding to the word, and a voice signal in which the user pronounces the word, extracts a speaker component corresponding to the word, and transmits it to a server Means for embedding in a document and sending it to the client; Means for receiving a speaker component corresponding to the word from the client; Means for extracting speech pattern data corresponding to the word by inputting the speaker component; Means for retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Means for determining whether the speech pattern data and the reference pattern data match; Means for sending a permission notification to the client if the voice pattern data and the reference pattern data match as a result of the determination; And And a means for transmitting a rejection notification to the client if the voice pattern data and the reference pattern data do not match, as a result of the determination. In the network-based speaker identification device for authenticating a user connected using a microphone, Means for receiving a member identifier from a client; Means for retrieving a word-group identifier stored in a web site database corresponding to the member identifier; Means for selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier; Means for embedding the word and a word identifier corresponding to the word in a first document to transmit to the client; Means for receiving a voice signal corresponding to the word from the client; Means for preprocessing the speech signal to extract a speaker component corresponding to the word; Means for extracting speech pattern data corresponding to the word by inputting the speaker component; Means for retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Means for determining whether the speech pattern data and the reference pattern data match; Means for sending a permission notification to the client if the voice pattern data and the reference pattern data match as a result of the determination; And And a means for transmitting a rejection notification to the client if the voice pattern data and the reference pattern data do not match, as a result of the determination. A network-based speaker learning apparatus for authenticating a user who is connected by using a microphone of a network server system among server systems, Means for receiving from the client predetermined input including at least a member identifier and storing it in a website database; Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group Means for storing in the web site database in conjunction with the details; Pre-processing a plurality of words belonging to the word group, a pre-built word identifier corresponding to each word, and a voice signal in which the user pronounces the words at the client to extract speaker components corresponding to each word Means for embedding component software in a first document for transmission to a server to transmit to the client; Means for receiving a member identifier, a word identifier corresponding to each word, and a speaker component corresponding to each word from the client and transmitting them to a speaker recognition server system; And And a means for receiving a learning completion notification from the speaker recognition server system and transmitting the learning completion notification to the client. The apparatus for learning a speaker using a network-based word group for user authentication in a network server system. A network-based speaker learning apparatus for authenticating a user who is connected by using a microphone of a speaker recognition system among server systems, Means for receiving a member identifier, word identifiers corresponding to each word, and speaker components entered directly from the client or via a network server system; Means for extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; And Means for generating a reference pattern identifier corresponding to each of the reference pattern data and storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker learning apparatus using network-based word group for user authentication in Chinese speaker recognition server system. A network-based speaker learning apparatus for authenticating a user who is connected by using a microphone of a network server system among server systems, Means for receiving from the client predetermined input including at least a member identifier and storing it in a website database; Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group Means for storing in the web site database in conjunction with the details; Means for embedding a plurality of words belonging to the word group and a pre-built word identifier corresponding to each word into a first document and transmitting the same to the client; Means for receiving a member identifier, a word identifier corresponding to each word and a voice signal corresponding to each word from the client and transmitting them to a speaker recognition server system; And And a means for receiving a learning completion notification from the speaker recognition server system and transmitting the learning completion notification to the client. The apparatus for learning a speaker using a network-based word group for user authentication in a network server system. A network-based speaker learning apparatus for authenticating a user who is connected by using a microphone of a speaker recognition system among server systems, Means for receiving a member identifier, word identifiers corresponding to each word, and voice signals input directly from the client or via a network server system; Means for preprocessing the speech signals to extract speaker components corresponding to each word; Means for extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; And Means for generating a reference pattern identifier corresponding to each of the reference pattern data and storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker learning apparatus using network-based word group for user authentication in Chinese speaker recognition server system. In the network-based speaker identification device for authenticating a user connected by using the microphone of the network server system of the server system, Means for receiving a member identifier from a client; Means for retrieving a word-group identifier stored in a web site database corresponding to the member identifier; Means for selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier; A component software for preprocessing the word, a word identifier corresponding to the word, and a voice signal in which the user pronounces the word in the client, extracting a speaker component corresponding to the word and transmitting it to the network server system ( means for embedding component software into a first document and sending it to the client; Means for receiving the speaker component from the client and transmitting the speaker component, the member identifier and the word identifier to a speaker recognition server system; Means for receiving a permission notification from the speaker recognition server system and transmitting it to the client; And And a means for receiving a rejection notification from the speaker recognition server system and transmitting the rejection notification to the client. A network-based speaker identification apparatus for authenticating a user who is connected by using a microphone of a speaker recognition system among server systems, Means for receiving a member identifier, word identifier and speaker component entered directly from a client or via a network server system; Means for extracting speech pattern data corresponding to the word by inputting the speaker component; Means for retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Means for determining whether the speech pattern data and the reference pattern data match; Means for sending a permission notification to the client or the network server system if the voice pattern data and the reference pattern data match as a result of the determination; And And a means for transmitting a rejection notification to the client or the network server system if the voice pattern data and the reference pattern data do not match as a result of the determination. Speaker verification apparatus using a network-based word group for. In the network-based speaker identification device for authenticating a user connected by using the microphone of the network server system of the server system, Means for receiving a member identifier from a client; Means for retrieving a word-group identifier stored in a web site database corresponding to the member identifier; Means for selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier; Means for embedding the word and a word identifier corresponding to the word in a first document to transmit to the client; Means for receiving a voice signal from the client and transmitting the voice signal, the member identifier and the word identifier to a speaker recognition server system; Means for receiving a permission notification from the speaker recognition server system and transmitting it to the client; And And a means for receiving a rejection notification from the speaker recognition server system and transmitting the rejection notification to the client. A network-based speaker identification apparatus for authenticating a user who is connected by using a microphone of a speaker recognition system among server systems, Means for receiving a member identifier, a word identifier and a voice signal input directly from the client or via a network server system; Means for preprocessing the speech signal to extract a speaker component corresponding to the word; Extracting voice pattern data corresponding to the word by using the speaker component as an input; Means for retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Means for determining whether the speech pattern data and the reference pattern data match; Means for sending a permission notification to the client or the network server system if the voice pattern data and the reference pattern data match as a result of the determination; And And a means for transmitting a rejection notification to the client or the network server system if the voice pattern data and the reference pattern data do not match as a result of the determination. Speaker verification apparatus using a network-based word group for. A memory storing a web site database, a voice information database, and a program; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Registering the user by storing user image information together with user voice characteristic information obtained by analyzing a voice signal of the user, Receiving a user authentication request from an authentication requester directly or indirectly through the network, Selecting a word or sentence corresponding to the state of the user from a pre-built word group or sentence group and transmitting the same to the authentication requester; Receiving the voice requestor's voice signal for the word or sentence or voice property information of the authentication requester extracted from the voice signal through the network; and Determining whether the voice signal or voice property information of an authentication requester extracted from the voice signal is input by the user using the registered user voice property information. User authentication system using a network-based word group, characterized in that to execute. A memory storing a web site database, a voice information database, and a program; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving from the client some input including at least a member identifier and storing it in a website database, Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group In conjunction with the information stored in the website database, Preprocessing a plurality of words belonging to the word group, a pre-built word identifier corresponding to each word, and voice signals in which the user corresponding to the client pronounces the words in the client to correspond to each word Extracting speaker components to be transmitted to the client by embedding component software for transmitting to the server to the client; Receiving word identifiers and speaker components corresponding to each word from the client, Extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; and Generating a reference pattern identifier corresponding to each of the reference pattern data, and then storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker learning system using network-based word group for authentication. A memory storing a web site database, a voice information database, and a program; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving from the client some input including at least a member identifier and storing it in a website database, Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group In conjunction with the information stored in the website database, Embedding a plurality of words belonging to the word group and a pre-built word identifier corresponding to each word in the first document and transmitting the same to the client; Receiving word identifiers and voice signals corresponding to each word from the client, Preprocessing the voice signals to extract speaker components corresponding to each word; Extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; and Generating a reference pattern identifier corresponding to each of the reference pattern data, and then storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker learning system using network-based word group for authentication. A memory storing a web site database, a voice information database, and a program; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving a member identifier from a client, Retrieving a word group identifier stored in a web site database corresponding to the member identifier; Selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier, A first component software for preprocessing the word, the word identifier corresponding to the word, and a voice signal in which the user pronounces the word, extracts a speaker component corresponding to the word, and transmits it to a server Embedding in a document and sending it to the client, Receiving a speaker component corresponding to the word from the client, Extracting voice pattern data corresponding to the word by inputting the speaker component; Retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Determining whether the voice pattern data and the reference pattern data match each other; Transmitting a permission notification to the client if the voice pattern data and the reference pattern data match as a result of the determination; And if the voice pattern data and the reference pattern data do not coincide with each other as a result of the determination, transmitting a rejection notification to the client. A memory storing a web site database, a voice information database, and a program; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving a member identifier from a client, Retrieving a word group identifier stored in a web site database corresponding to the member identifier; Selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier, Embedding the word and a word identifier corresponding to the word in a first document and transmitting the same to the client; Receiving a voice signal corresponding to the word from the client, Preprocessing the voice signal to extract a speaker component corresponding to the word; Extracting voice pattern data corresponding to the word by inputting the speaker component; Retrieving reference pattern data stored in a voice information database corresponding to the word identifier, Determining whether the voice pattern data and the reference pattern data match each other; Transmitting a permission notification to the client if the voice pattern data and the reference pattern data match as a result of the determination; And if the voice pattern data and the reference pattern data do not match, determining to send a rejection notification to the client. Memory that stores Web site databases and programs; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving from the client some input including at least a member identifier and storing it in a website database, Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group In conjunction with the information stored in the website database, Preprocessing a plurality of words belonging to the word group, a pre-built word identifier corresponding to each word, and a voice signal in which the user corresponding to the client pronounces the words in the client to correspond to each word Extracting speaker components to be transmitted to the client by embedding component software for transmitting to the server to the client; Receiving a member identifier, a word identifier corresponding to each word, and a speaker component corresponding to each word from the client and transmitting them to a speaker recognition server system; And receiving a learning completion notification from the speaker recognition server system and transmitting the learning completion notification to the client. The speaker learning network server system using the network-based word group for user authentication. A memory storing a voice information database and a program; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving a member identifier, word identifiers corresponding to each word, and speaker components entered directly from a client or via a network server system, Extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; and Generating a reference pattern identifier corresponding to each of the reference pattern data, and storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker Recognition Speaker Recognition Server System using Network-based Word Group for Users. Memory that stores Web site databases and programs; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving from the client some input including at least a member identifier and storing it in a website database, Determine a word group among a plurality of word groups pre-built in the web site database, wherein each word group consists of a plurality of words, and input a pre-built word group identifier corresponding to the word group In conjunction with the information stored in the website database, Embedding a plurality of words belonging to the word group and a pre-built word identifier corresponding to each word in the first document and transmitting the same to the client; Receiving a member identifier, a word identifier corresponding to each word, and a voice signal corresponding to each word from the client and transmitting them to a speaker recognition server system; and And receiving a learning completion notification from the speaker recognition server system and transmitting the learning completion notification to the client. The speaker learning network server system using the network-based word group for user authentication. A memory storing a voice information database and a program; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving a member identifier, word identifiers corresponding to each word and voice signals input directly from the client or via a network server system, Preprocessing the voice signals to extract speaker components corresponding to each word; Extracting reference pattern data corresponding to each word by performing a learning process on the speaker components; and Generating a reference pattern identifier corresponding to each of the reference pattern data, and storing the member identifier, the reference pattern data, the reference pattern identifier, and a word identifier in a voice information database. Speaker Recognition Speaker Recognition Server System using Network-based Word Group for Users. Memory that stores Web site databases and programs; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving a member identifier from a client, Retrieving a word group identifier stored in a web site database corresponding to the member identifier; Selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier, Preprocessing the word, a word identifier corresponding to the word, and a voice signal in which the user corresponding to the client pronounces the word in the client extracts a speaker component corresponding to the word and transmits it to the network server system. Embedding component software in a first document for transmission to the client, Receiving a speaker component from the client and transmitting the speaker component, the member identifier and the word identifier to a speaker recognition server system; Receiving a permission notification from the speaker recognition server system and sending it to the client; and And receiving the rejection notification from the speaker recognition server system and transmitting the rejection notification to the client. A memory storing a voice information database and a program; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving a member identifier, word identifier and speaker component entered directly from a client or via a network server system, Extracting voice pattern data corresponding to the word by inputting the speaker component; Retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Determining whether the voice pattern data and the reference pattern data match each other; Transmitting a permission notification to the client or the network server system when the voice pattern data and the reference pattern data match with the determination result; If the voice pattern data and the reference pattern data do not match, a step of transmitting a rejection notification to the client or the network server system, wherein the speaker uses the network-based word group for user authentication. Check speaker recognition server system. Memory that stores Web site databases and programs; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving a member identifier from a client; Retrieving a word group identifier stored in a web site database corresponding to the member identifier; Selecting one word among a plurality of words belonging to a word group corresponding to the word group identifier, Embedding the word and a word identifier corresponding to the word in a first document and transmitting the same to the client; Receiving a voice signal from the client and transmitting the voice signal, the member identifier and the word identifier to a speaker recognition server system; Receiving a permission notification from the speaker recognition server system and sending it to the client; and And receiving the rejection notification from the speaker recognition server system and transmitting the rejection notification to the client. A memory storing a voice information database and a program; A processor coupled to the memory to execute the program Including but not limited to: The processor by the program, Receiving a member identifier, a word identifier and a voice signal input directly from the client or via a network server system, Preprocessing the voice signal to extract a speaker component corresponding to the word; Extracting voice pattern data corresponding to the word by inputting the speaker component; Retrieving reference pattern data stored in a voice information database corresponding to the word identifier; Determining whether the voice pattern data and the reference pattern data match each other; Transmitting a permission notification to the client or the network server system when the voice pattern data and the reference pattern data match with the determination result; If the voice pattern data and the reference pattern data do not match, a step of transmitting a rejection notification to the client or the network server system, wherein the speaker uses the network-based word group for user authentication. Check speaker recognition server system. A method of restoring voice information for preventing a customer from arranging or denying a contract or an order after a contract is established by a customer in at least one order or subscription in an e-commerce, Transmitting a word or sentence corresponding to each order or subscription from the pre-stored word group to the client; Receiving a purchase order voice signal corresponding to the word or sentence from the client; Storing voice pattern data corresponding to the purchase order voice signal in a voice information database; Voice information restoration method using a word group comprising a. 92. The method of claim 92, The voice information restoration method using the word group Restoring the voice pattern data to the purchase order voice signal when the customer establishes or denies the fact of the contract establishment or order. Speech information restoration method using a word group characterized in that it further comprises. 92. The method of claim 92, Voice information restoration method using the word group, Receiving the purchase order voice signal from a client; Preprocessing the purchase order voice signal to extract a speaker component corresponding to the purchase order voice signal; And Extracting speech pattern data corresponding to the purchase order speech signal from the speaker component Speech information restoration method using a word group characterized in that it further comprises.