KR20010110964A - The method for verifying users by using voice recognition on the internet and the system thereof - Google Patents

Abstract

The present invention is to ensure the security of user information at low cost by using voice for authentication of Internet service users on the Internet.

The method for authenticating a user using voice on the Internet of the present invention includes the steps of registering a user's voice, receiving a user's voice for user authentication, and using the voice received from the user. Comparing the result with the voice, and storing the result of the comparison in a database.

Description

The method for verifying users by using voice recognition on the internet and the system approximately}

The present invention relates to a method and system for registering and authenticating a user using voice, and more particularly, to a method and system for registering and authenticating an Internet service user using voice over the Internet.

The class that uses the Internet has already spread out to a certain number of groups to the public, and the Internet market is a place with a lot of room for expansion and expansion, unlike the already saturated market. In addition, security and authentication are becoming more important in electronic commerce through a network.

Until now, most users have used the ID / Password method using a database to authenticate users on a network such as the Internet. In the ID / Password method, when an Internet service user accesses the initial site of the Internet service provider and inputs his or her ID and password, the Internet service provider stores the ID and password of the Internet service user in the database of the system. The user is authenticated according to the ID and password of the database user.

In addition, as a user authentication method which is of recent interest, when an Internet service user or the like submits to an Internet service provider using a certificate issued from a certification authority, the Internet service provider reads the submitted certificate and authenticates the user. .

In the conventional ID / PASSWORD method described above, the impersonator can easily use another person's ID and PASSWORD by stealing ID and PASSWORD. Therefore, this may not be a big security problem for a user using a financial system such as a bank. none.

On the other hand, biometric information such as fingerprints, voices, and faces maintain their own unique forms, and they can be used for identity verification and identity verification because they are not lost or forgotten. Among them, the voice recognition method is used for recognition. The hardware used has advantages in terms of cost since it can use existing widely used equipment as it is.

In view of the above, the present invention is to ensure the security of user information at low cost by registering and authenticating a user using voice over a network such as the Internet.

1 is a schematic diagram of one embodiment of a network system on the Internet of the present invention;

2 is a block diagram of a voice authentication server of the present invention.

3 is a flowchart showing a process for registering a user's voice;

4 is a block diagram showing an initial screen for accessing an Internet for a user to register a voice using the voice authentication system of the present invention.

5 is a block diagram showing a process in which a user's voice is registered in a speaker DB in the voice authentication system of the present invention.

6 is a flowchart showing a process for authenticating a user's voice.

FIG. 7 is a block diagram showing an initial screen of an Internet connection for authenticating a voice by a user using the voice authentication system of the present invention. FIG.

8 is a block diagram showing a process of authenticating a user's voice in the voice authentication system of the present invention.

9 is a schematic diagram of another embodiment of a network system on the Internet in which the voice authentication system of the present invention is used.

The authentication system of the present invention can be divided into a voice authentication service provider side and a voice authentication service user side. A voice authentication service provider includes a voice authentication server performing a voice recognition / speaker authentication function, and a voice authentication server and a voice authentication service user side. It includes a relay module for communication relay of, the voice authentication service user includes an Internet service user and an Internet service provider.

The voice authentication server performs a voice recognition function, a speaker registration function, and a speaker authentication function. The purpose of performing voice recognition is to prevent impersonators who randomly recorded a user's voice, for example, inputting a predetermined number of numeric sounds. If the number to be entered matches the requested number, the voice recognition is successful (voice recognition). When the voice recognition succeeds in this way, the feature parameters of the user's individual voice requesting user registration are stored and registered in the speaker DB (speaker registration). After the authentication request is sent from the user, as in the speaker registration, if the voice recognition is successful after the voice recognition, the similarity is obtained by using the characteristics of the input user's voice and the feature parameters of the voice stored in the speaker DB. When the similarity obtained is compared with a predetermined threshold and the similarity is greater than or equal to the threshold value, the speaker authentication succeeds, and when the similarity level is less than the threshold value, the speaker authentication fails (speaker authentication or voice authentication).

One feature of the present invention relates to a method for authenticating a user using voice over the Internet, the method comprising: registering a user's voice, receiving a user's voice desired for user authentication, and receiving from the user Comparing the registered voice with the registered voice, and storing the comparison result in a database.

According to another aspect of the present invention, in the method for authenticating a user by using voice on the Internet, the step of receiving a voice of the user who wants to authenticate the user is predetermined to a user who wants to log on to an Internet service of an Internet service provider. Requesting utterance, and receiving a voice of a user who wishes to authenticate a user for logging on to the Internet service, and controlling the Internet service logon of the Internet service provider by the user according to the result stored in the database. It further includes the steps.

Still another aspect of the present invention provides a method of authenticating a user using voice on the Internet, the method further comprising: requesting a user who desires user authentication to have a predetermined voice, using the voice received from the user. The comparing with the registered voice may include determining whether the voice received from the user is a predetermined utterance requested by the user, and if the voice received from the user is not determined to be the predetermined utterance requested by the user, Requesting a predetermined utterance again, and if it is determined that the voice received from the user is the requested utterance, comparing the voice received from the user with the registered voice.

In still another aspect of the present invention, there is provided a method of authenticating a user using voice on the Internet, including registering an ID and a password of a user, and receiving an ID and a password of a user who wants to log on to an internet service of an internet service provider. And receiving the voice of the user who wishes to log on to the internet service of the internet service provider, only if the received user ID and password match the registered ID and password. To receive your voice.

According to yet another aspect of the present invention, in the method of authenticating a user using voice on the Internet, controlling the Internet service logon of the Internet service provider by the user according to the comparison result stored in the database, the comparison result Transmitting the data to an Internet service provider for which the user wishes to log on.

Another aspect of the present invention relates to a method for controlling a user's logon on an Internet service using voice over the Internet, the method comprising: registering a user's voice, receiving a user's voice for an Internet service, Comparing the registered voice with the registered voice by using the received voice, and controlling the logon of the Internet service by the user according to the comparison result.

Another aspect of the present invention relates to a system for authenticating a user using voice over the Internet, comprising: a basic information database storing a user ID and a password, a speaker database storing the characteristics of a user's voice, and voice recognition. A storage device including a voice information database used for storing the information, an authentication result database storing the authentication result, and requesting a user to speak a speaker, and receiving a voice from the user. A voice recognition module for determining whether a voice is a predetermined voice, and when the voice received from the user determines that the voice received from the user is a requested voice, the characteristics of the voice received from the user and the voice stored in the speaker database are similar to each other. Determines if the threshold exceeds And the result of the determination includes a speaker authentication module to store the authentication result module.

Another feature of the present invention is a system for registering and authenticating a user using voice on the Internet, wherein the communication service with the Internet service provider and the user further includes the relay module, which is made through the relay module.

Hereinafter, the present invention will be described in detail with reference to the drawings.

1 is a configuration of one embodiment of a network system according to the present invention, a voice authentication service provider 10 including a voice authentication server 100 and a relay module 101, an Internet service user (hereinafter, 'user', 102), an Internet service provider (hereinafter, 'provider', 103). The voice authentication server, Internet service user, and Internet service provider are connected through a communication network. Here, the network means the Internet, but other network networks, for example, a LAN, a WAN, a Public Switched Telephone Network (PSTN), a Public Switched Data Network (PSDN), a cable TV network, a wireless communication network, and the like, are also possible. The voice authentication server 100 is a server that registers a user's voice in response to a request from a user who wishes to use the voice authentication system and provides an authentication service when there is an authentication request from the user. In addition, the communication between the Internet service user or the Internet service provider and the voice authentication server 100 is always performed by the relay module 101 to prevent direct access to the voice authentication server 100 to prevent unauthorized intrusion of the Internet service user and the Internet service provider. Through). The Internet service user 102 transmits his or her personal information and voice through the communication network to register the user, and when voice authentication is desired, transmits the voice requested by the voice authentication server and receives the authentication result. The Internet service provider 103 receives the personal information and the authentication result of the Internet service user from the voice authentication server when the authentication result of the voice authentication server is successful.

2 is a block diagram schematically showing the configuration of the voice authentication service provider 10 of the present invention. The voice authentication server 100 includes a voice recognition control module 201, a speaker registration control module 202, a control module including a speaker authentication control module 203, a user's name, social security number, ID, password, and the like. A user personal information DB 204 for storing information, a voice information DB 205 for storing general pattern information of a voice for a voice recognition function, a speaker DB 206 for storing feature parameters extracted from a user's voice, and , A number of functions used for speech recognition, speaker registration, speaker authentication (e.g., functions for receiving data from the user, functions for generating and transmitting all messages to the user, functions related to parameter extraction of the user's voice, user A voice DB management function, a voice authentication algorithm related function), and a function DB 207 for storing the voice authentication result of the user.

When the voice recognition control module 201 receives a voice from the Internet service user, the voice recognition control module 201 determines whether the received voice is the voice requested by the voice authentication server, and when the voice is requested, the voice recognition control module 201 is sent to the next processing module such as speaker registration or speaker authentication. If the data is transmitted, and the requested voice is not the requested voice, the voice is requested again from the user or the voice recognition result is output.

The speaker registration control module 202 controls to extract the characteristic parameter of the user's individual from the voice of the user whose voice recognition is successful and store it in the speaker-specific DB 206. The speaker authentication control module 203 controls the authentication of the voice of the user requesting voice authentication using the data stored in the speaker-specific DB 206.

The relay module 101 relays the data output from the voice authentication server 100 and transmits the data to the Internet service user and the Internet service provider through the network, and receives the data from the Internet service user through the network to the voice authentication server 100. To be sent). In this way, the relay module 101 relays the communication between the voice authentication server and the Internet service user. In this case, the relay module 101 checks whether each packet is a message format defined by the authentication system and discards the packet if it is not a defined packet. In addition, when the system is composed of a plurality of voice authentication server, the relay module may distribute the communication connection with the Internet service users according to the load of the voice authentication server.

Hereinafter, a speaker registration process for registering a user's voice will be described with reference to FIGS. 3 to 5, and a speaker authentication process for authenticating a speaker using the registered voice will be described with reference to FIGS. 6 to 8.

3 is a flowchart for speaker registration. The voice authentication server receives a speaker registration request from the user (step 301) and requests the user to enter the user's personal information (step 302). Next, the voice authentication server receives the personal information input by the user (step 303), stores the personal information in the user personal information DB, and requests the user for voice data to learn (step 304). Next, the voice authentication server receives voice data from the user (step 305), and determines whether the voice data received from the user is voice data requested by the voice authentication server using the voice recognition function (step 306). The voice recognition function will be described in detail with reference to FIG. 5. In the case of the recognition failure as a result of the determination of the voice authentication server, that is, when it is determined that the speech data received from the user is not the requested speech data, the user is asked again to speak the speech data to be learned (step 304). If it is determined that the voice data received from the user is the requested voice data, the feature parameter is extracted from the received voice data of the user (step 307), and the extracted feature parameter is stored in the speaker-specific DB 206 (step 308). ). In this way, the voice authentication server extracts and stores features from the user's voice to complete speaker registration.

An initial screen in which the user can input personal information and voice data is shown in FIG. 4 to explain the process (steps 302 to 305) of receiving the user's personal information and voice data. When the user who wishes to register a speaker connects to the voice authentication server, the initial screen of Fig. 4 is displayed on the display device of the user computer. In this embodiment, the user directly connects to the voice authentication server as an example, but after accessing the web page of the desired Internet service provider, the initial screen of the voice authentication server as shown in FIG. 4 is connected to the user. May be shown to The user enters data in a form that allows the user to enter an ID, password, name, social security number, and account number included in the user information input menu. After the user inputs all personal information, the user inputs his or her own voice by clicking the recording icon included in the voice input menu. Follow the instructions at the top of the voice input menu to read the numbers 0 to 9 three times, and when the voice input is finished, click the stop icon to end recording. The play icon can be used when the user wishes to hear his / her recorded voice again. After the user inputs both the personal information and the voice, by clicking the end icon displayed on the upper right of the screen, the personal information data and the voice data input by the user may be transmitted to the voice authentication server through the relay module 101. In fact, when recording is completed, wave-type data is first generated, and when raw data is successfully generated from which header information of the DC component and the silence section wave-type data is successfully generated, raw data representing personal information data and voice information is generated. Is sent.

At the bottom of the voice input menu, a status bar indicating the status of the current operation is displayed. For example, when the voice authentication server receiving the user's personal information and voice information recognizes the voice recognition result, the status bar is displayed. In the case of "voice input succeeded" or in case of recognition failure, the message "voice input failed" may be displayed in the status bar. In the case of recognition failure, the user repeats the voice input process again and transmits the user's own voice data to the voice authentication server 100 through the relay module.

Next, a process of registering the voice received from the user with the speaker DB will be described with reference to FIG. 5. The preprocessing process 503 and the similarity measurement 504 of FIG. 5 are parts for performing the user's speech recognition, and the pattern learning process 505 is for performing speaker registration for registering the user's voice in the speaker DB 207. to be. The preprocessing process 501 and the pattern learning process 502 are for showing a process of constructing the voice information DB 204 used for speech recognition.

First, the preprocessing process 501 is a process of receiving a sound signal to obtain a feature of the voice, detecting the start point and the end point of the input sound signal to remove the non-signal part, and find the voice part (Endpoint Detection), Endpoint A speech frame is divided by segmenting the input speech signal from the detection, and the feature value for the interval is found (LPC-Cep (linear prediction coefficient 계수 strum) feature extraction). The speech feature extraction portion is extracted in the order of pre-emphasis, Hamming windows, and linear predictive coefficients (LPC) and then LPC cepstrum. The pattern learning process 502 is a process of creating a probabilistic model using the feature values of the speech obtained in the preprocessing process. The pattern learning process 502 is initialized to use the CHMM algorithm for establishing the probabilistic model of the speech (Continuous Hidden Markov Model (CHMM) model). Initialization), calculates the probability model of speech using CHMM (Training process), checks whether the calculated probability model converges and is stable, and if convergence is confirmed, CHMM model parameter λ (π, A, C of speech information) ,?,?) are stored in the voice information DB 204.

Now, when the voice data of the user who wants to register the speaker is input to the preprocessing process 503, the voice part of the input data is found (Endpoint Detection), and the feature value of the voice is extracted from the voice signal (LPC-Cep characteristic). extraction). The similarity measure 504 measures the degree to which the input voice signal is similar using a previously established probability model. The similarity measure 504 stores the feature parameter X (t) received from the preprocessing 503 and the feature stored in the voice information DB 204. It is to determine how well the standard pattern of the parameter λ matches. At this time, for example, if λ stored in the voice information DB corresponds to the numeric tones 0 to 9, X (t) is matched to the numeric tones 0 to 9, respectively, with the largest similarity Max (arg Likelihood). Perform voice recognition.

For example, if speech input for "3" is performed and the input speech is "4", Max (arg Likelihood) for input speech "4" will appear at 4, which is speech recognition as "3". No, voice recognition is a failure.

If the voice recognition fails by the above process, the voice data is input again from the user to measure similarity with the preprocessing process. If the voice recognition is successful, the pattern learning process is used to obtain information to be used for speaker authentication from the inputted user's voice. 505 is performed. The pattern learning process 505 is a process of creating a probabilistic model using the feature values of the speech obtained in the preprocessing process 503. Since the pattern learning process 505 creates a probabilistic model in which the feature information of the user's own speech is well represented, There is a difference in the model from the pattern learning process 502, which is a process of creating a probabilistic model containing characteristic information of general speech. In the pattern learning process 505, an initialization is performed to use a GMM algorithm for establishing a probability model of speech (GMM (Gaussian Mixture Model) model initialization), and a probability model of the speech is calculated using the GMM (Training process). The purpose of the training process is to estimate the parameter λ of the GMM which, when given training speech from the speaker, maps the distribution of the feature vector well. Next, it is checked whether the calculated probability model converges and is stable.When convergence is confirmed, the GMM model parameters λ (p, μ, Σ) (where p is the mixture weight, μ is the mean vector, and Σ is A covariance matrix) is stored in the speaker DB 206. By this process, the characteristics of the user's personal voice are registered in the speaker DB 207.

6 is a flowchart for speaker authentication. The voice authentication server receives a speaker authentication request from the user (step 601) and requests the user a predetermined voice (step 602). The predetermined voice requested by the voice authentication server may be a character sound or a numeric sound, and the voice or numeric sound requested by the voice authentication server to prevent impersonation by recording the user's voice registered in the speaker DB. Is combined by a predetermined random number. The voice authentication server receives the user's voice (step 603) and extracts feature parameters from the received user's voice data (step 604). Next, it is determined whether the speech received from the user is the requested speech data using voice recognition (step 605), and in case of recognition failure, the user is asked for a predetermined speech again, and in case of recognition success, the extracted feature parameter. And the feature parameters stored in the speaker DB 206 to determine whether there is a match, and if the authentication fails, a "authentication failure" message is displayed on the display device of the user computer. Send personal information and authentication result to internet service provider, and display "successful authentication" message to internet service user. By this process, speaker authentication is performed.

Fig. 7 shows an initial screen for inputting voice data so that a user who has registered a speaker can request speaker authentication. When a user who wishes to authenticate a speaker connects to the voice authentication server, an initial screen as shown in Fig. 7 is displayed on the display device of the user computer. As in the case of speaker registration, the user who wishes to authenticate the speaker may access the initial screen of the voice authentication server as shown in FIG. 7 after accessing the web page of the desired Internet service provider. The user inputs his or her ID and PASSWORD in the form provided so that the user can enter the ID and PASSWORD included in the user input menu. Next, the user clicks the recording icon included in the voice input menu, reads the number (***) displayed in the voice input menu, and clicks the stop icon when voice input is completed. Next, when the user clicks the authentication icon, the ID, password, and user's voice input by the user are transmitted to the voice authentication server 100 through the relay module through the network.

8 illustrates a process in which the voice authentication server that receives the voice of the user requesting the speaker authentication performs the speaker authentication. The user's voice recognition is performed in the similarity measure 802, and the speaker authentication is performed in the test process 803.

When the user's voice S (n) is received, the user's voice recognition is performed by the same method as in the preprocessing 503 and the similarity measurement 504 of FIG. In addition, if voice recognition fails, the user is asked to speak again, and if successful, a test process 803 for speaker authentication is performed. In the test process 803, the similarity is obtained by inputting and matching the feature parameter X (t) of the user's voice obtained in the preprocessing process 801 and the GMM model parameter λ (p, μ, Σ) stored in the speaker DB DB 206. The similarity is compared with a threshold value (a threshold value that can be determined to be included in the GMM model stored in the speaker DB) obtained from the speaker registration process (compared with the threshold). As a result of the comparison, if the similarity value is equal to or higher than the threshold, the speaker is considered to be a match, and if it is lower than the threshold, the speaker is classified as a success. The result of the speaker authentication as described above is stored in the authentication result DB 207.

According to the authentication result stored in the authentication result DB according to the above method, the message "Authentication Success" or "Authentication Failure" is transmitted to the Internet service user through the relay module and the network, and the authentication result is used by the Internet service user. Can be sent to an Internet service provider that wishes to: One specific example of transmitting the authentication result to the Internet service user and the Internet service provider is as follows. After performing the speaker authentication, the voice authentication server sends a result to the user in the form of a packet. If the authentication result is successful, the voice authentication server is connected to the web page of the Internet service provider. On the other hand, the voice authentication server stores the user's ID, the authentication result storage time, the user's IP address, etc. in the authentication result DB only when the authentication result is successful, and the user transmits the authentication result of the voice authentication server to the Internet browser. When accessing the service provider's web page, the Internet service provider queries the voice authentication server using the user's IP address to search the authentication result DB, and searches and matches the stored IP address with the user's IP address. If it is, the stored time is checked, and if the difference between the stored time and the current time is, for example, within 30 seconds, the user is allowed to log on, otherwise the access logon is blocked. In addition, the voice authentication server deletes all authentication results over 30 seconds from the authentication result DB, regardless of the connection permission failure.

9 shows another embodiment of a network system on the Internet in which the voice authentication system of the present invention is used. Unlike the embodiment shown in FIG. 1, in the embodiment of FIG. 9, the Internet service provider 103 includes a voice authentication server 100 that performs voice authentication. Therefore, when the Internet service provider 103 receives the voice authentication request from the Internet service user 102 through the network, the voice service is performed by the voice authentication server 100 and then the Internet service is performed using the voice authentication result. You can allow or prohibit a user from logging on to access. In this embodiment, since the Internet service provider 103 includes the authentication result of the voice authentication server, the authentication result of the voice authentication server is transmitted to the Internet service provider 103 through the network as shown in the embodiment shown in FIG. There is no need.

Each of the above methods may be reversed in order to perform this method. In addition, various changes and modifications may be made by those skilled in the art in the field to which the present invention pertains, and the present invention is not limited to the examples described above, but is defined by the scope described in the claims.

According to the present invention, by registering and authenticating a user using voice over a network such as the Internet, the user information can be secured without incurring a high cost.

Claims (8)

In the method of authenticating a user using voice on the Internet, Registering the user's voice; Receiving a voice of a user who desires user authentication, Comparing the registered voice with the registered voice using the voice received from the user; Storing the result of the comparison in a database. The method of claim 1, Receiving the voice of the user who wants to authenticate the user, Requesting a predetermined voice from a user who wishes to log on to the Internet service of the Internet service provider; Receiving a voice of a user who wishes to authenticate a user for logging on to the Internet service, Controlling the logon of the internet service provider by the user according to the result stored in the database. The method of claim 1, Requesting a predetermined voice from a user who desires user authentication, Comparing with the registered voice using the voice received from the user, Determining whether the voice received from the user is a predetermined voice requested by the user; As a result of the determination, if the voice received from the user is not judged to be the predetermined utterance requested, the user requests the predetermined utterance again, and if the voice received from the user is determined to be the requested utterance, the voice received from the user And comparing the registered voice with the registered voice. The method of claim 1, Registering the user's ID and password, Receiving an ID and PASSWORD of the user who wants to log on to the Internet service of the Internet service provider, Receiving a voice of a user who wants to log on to the Internet service of the Internet service provider, And receiving the user's voice only when the received user's ID and password match the registered ID and password. The method of claim 2, Controlling the Internet service logon of the Internet service provider by the user according to the comparison result stored in the database includes transmitting the comparison result to an Internet service provider to which the user wishes to log on. To authenticate the user using a. A method of controlling a user's logon on an Internet service using voice on the Internet, Registering the user's voice; Receiving a voice of a user who desires an Internet service, Comparing the registered voice with the registered voice using the received voice; Controlling an Internet service logon by the user according to the comparison result. In a system for authenticating a user using voice on the Internet, A storage device including a basic information database storing user ID and password, a speaker database storing user's voice characteristics, a voice information database used for voice recognition, and an authentication result database storing authentication results; , A voice recognition module for requesting a speaker to request a speaker, and for determining whether the received voice is a predetermined voice requested by receiving a voice from the user; If the voice recognition module determines that the voice received from the user is the requested utterance, it is determined whether the characteristics of the voice received from the user and the voice stored in the speaker database have similarities exceeding a predetermined threshold value. A system for registering and authenticating a user using voice on the Internet including a speaker authentication module for storing the determination result in the authentication result module. The method of claim 7, wherein A communication system for registering and authenticating a user using voice on the Internet, further comprising the relay module, wherein both the communication with the Internet service provider and the user is via a relay module.