KR102621881B1 - Server and method for authenticating speaker during call to counselor - Google Patents

Abstract

The speaker authentication server that authenticates the speaker consulting with the counselor includes a real-time speech acquisition unit that acquires the speaker's real-time real-time speech from the speaker's real-time consultation voice information, a real-time speech transmission unit that transmits the real-time real-time speech of the speaker to the score calculation server, and An authentication result receiving unit that receives a real-time authentication score for the speaker's real-time authentication result based on comparison between the speaker's real-time real-time voice at each set authentication cycle and the speaker's pre-registered voice fingerprint from the score calculation server, and based on the received authentication score It includes a risk determination unit that determines the risk of impersonation of the speaker, and the authentication cycle can be varied based on the real-time authentication score.

Description

Server and method for authenticating a speaker consulting with a counselor {SERVER AND METHOD FOR AUTHENTICATING SPEAKER DURING CALL TO COUNSELOR}

The present invention relates to a server and method for authenticating a speaker consulting with a counselor.

According to a conventional telephone consultation system, a counselor requests information about the customer (eg, the customer's name, date of birth, address information, payment method, credit card number, etc.) to identify and authenticate the customer.

Since these telephone consultation systems identify and authenticate customers only based on the information provided by the customer, there was a problem in which impersonators who illegally obtained information about the customer stole the information about the customer and used the service.

To solve the above-mentioned problems, a speaker authentication system that verifies identity using the customer's voice was recently introduced.

The speaker authentication system, which verifies identity using the customer's voice, registers the customer's voice for that specific sentence when the customer utters a specific preset sentence, and then combines the registered customer's voice with the input customer's voice. Authenticate customers by comparing.

However, since this speaker authentication system authenticates the customer through the customer's voice entered only at the start of the consultation, it is difficult to authenticate the customer in real time, and it is difficult to fundamentally block customer impersonation, as well as when there are noises from everyday life and changes in the customer's voice. There is a problem that it is difficult to authenticate the customer through the customer's voice.

Korean Patent Publication No. 2017-0105034 (published on September 18, 2017)

The present invention is intended to solve the problems of the prior art described above, and provides a real-time authentication score for the real-time authentication result of the speaker by comparing the real-time actual voice of the speaker consulting with the counselor at each preset authentication cycle and the speaker's pre-registered voice fingerprint. We want to calculate .

In addition, the present invention seeks to change the authentication cycle that determines the risk of impersonation of a speaker based on the authentication score and compares the speaker's real-time real-time voice and the speaker's pre-registered voice fingerprint based on the real-time authentication score. However, the technical challenges that this embodiment aims to achieve are not limited to the technical challenges described above, and other technical challenges may exist.

As a technical means for achieving the above-described technical problem, the speaker authentication server for authenticating a speaker consulting with a counselor according to the first aspect of the present invention includes a device for obtaining real-time real-time speech quality of the speaker from real-time consultation voice information of the speaker. Voice acquisition unit; a real-time speech transmission unit that transmits the real-time real-time speech of the speaker to a score calculation server; an authentication result receiving unit that receives, from the score calculation server, a real-time authentication score for the real-time authentication result of the speaker based on comparison between the real-time real-time voice of the speaker at each preset authentication cycle and the speaker's pre-registered voice fingerprint; and a risk determination unit that determines a risk of impersonation for the speaker based on the received real-time authentication score, and the authentication cycle can be varied based on the real-time authentication score.

A score calculation server that authenticates a speaker consulting with a counselor according to a second aspect of the present invention includes a real-time speech reception unit that receives real-time real-time speech of the speaker from a speaker authentication server; a real-time authentication unit that performs real-time authentication of the speaker at preset authentication cycles by comparing the speaker's real-time real-time voice quality with the speaker's pre-registered voice fingerprint; an authentication score calculation unit that calculates a real-time authentication score for the result of real-time authentication of the speaker; and an authentication score transmission unit that transmits the real-time authentication score to the speaker authentication server, wherein the real-time real-time voice of the speaker is obtained from the real-time counseling voice information of the speaker, and the authentication period is variable based on the real-time authentication score. It can be.

A method of authenticating a speaker consulting with a counselor in a speaker authentication server according to a third aspect of the present invention includes the steps of obtaining real-time real-time speech quality of the speaker from real-time counseling voice information of the speaker; Transmitting the real-time real-time speech of the speaker to a score calculation server; Receiving a real-time authentication score for the real-time authentication result of the speaker from the score calculation server based on comparison between the real-time real-time voice of the speaker at each preset authentication cycle and the speaker's pre-registered voice fingerprint; and determining a risk of impersonation for the speaker based on the received real-time authentication score, wherein the authentication cycle can be varied based on the real-time authentication score.

The above-described means for solving the problem are merely illustrative and should not be construed as limiting the present invention. In addition to the exemplary embodiments described above, there may be additional embodiments described in the drawings and detailed description of the invention.

According to one of the above-described means for solving the problem of the present invention, the present invention compares the real-time real-time voice of the speaker consulting with the counselor at each preset authentication cycle with the speaker's pre-registered voice fingerprint to provide real-time information on the real-time authentication result of the speaker. The certification score can be calculated.

In addition, the present invention can determine the risk of impersonation for a speaker based on the authentication score, and change the authentication cycle to compare the speaker's real-time real-time voice and the speaker's pre-registered voice fingerprint based on the real-time authentication score.

Through this, the present invention enables real-time speaker authentication from the start of the consultation to the end of the consultation, thereby ensuring the speaker's identity during the entire consultation process and fundamentally blocking impersonation. In other words, the present invention can fundamentally block impersonation that may occur due to a change in speaker during the consultation even if the speaker's identity is successfully verified at the start of the consultation.

In addition, the present invention can prevent impersonators from impersonating any speaker because speaker authentication is performed only through real-time conversation between the counselor and the speaker (i.e., real-time counseling voice information).

Additionally, the present invention can reduce the counselor's response time required to verify the speaker's identity. In addition, the present invention can reduce the resistance that previously occurred by providing the speaker's personal information during consultation with a counselor and improve the efficiency of counseling work.

In addition, the present invention can prevent the speaker's important personal information from being leaked, and can improve customer satisfaction by shortening the time required to verify identity during a consultation call.

1 is a configuration diagram of a speaker authentication system according to an embodiment of the present invention.
Figure 2 is a block diagram of the speaker authentication server shown in Figure 1, according to an embodiment of the present invention.
3A to 3D are diagrams for explaining a method of authenticating a speaker according to an embodiment of the present invention.
Figure 4 is a flowchart showing a method of authenticating a speaker in a speaker authentication server, according to an embodiment of the present invention.
Figure 5 is a block diagram of the score calculation server shown in Figure 1, according to an embodiment of the present invention.
Figure 6 is a flowchart showing a method of authenticating a speaker in a score calculation server according to an embodiment of the present invention.

Below, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. In order to clearly explain the present invention in the drawings, parts that are not related to the description are omitted, and similar parts are given similar reference numerals throughout the specification.

Throughout the specification, when a part is said to be "connected" to another part, this includes not only the case where it is "directly connected," but also the case where it is "electrically connected" with another element in between. . Additionally, when a part "includes" a certain component, this means that it may further include other components rather than excluding other components, unless specifically stated to the contrary.

In this specification, 'part' includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Additionally, one unit may be realized using two or more pieces of hardware, and two or more units may be realized using one piece of hardware.

In this specification, some of the operations or functions described as being performed by a terminal or device may instead be performed on a server connected to the terminal or device. Likewise, some of the operations or functions described as being performed by the server may also be performed on a terminal or device connected to the server.

Hereinafter, specific details for implementing the present invention will be described with reference to the attached configuration diagram or processing flow diagram.

1 is a configuration diagram of a speaker authentication system according to an embodiment of the present invention.

Referring to FIG. 1, the speaker authentication system may include a speaker authentication server 100 and a score calculation server 110. However, since the speaker authentication system of FIG. 1 is only one embodiment of the present invention, the present invention is not limited to FIG. 1, and may be configured differently from FIG. 1 according to various embodiments of the present invention.

In general, each component of the speaker authentication system of FIG. 1 is connected through a network (not shown). Network refers to a connection structure that allows information exchange between nodes such as terminals and servers, including Local Area Network (LAN), Wide Area Network (WAN), and World Wide Area Network (WWW). Wide Web), wired and wireless data communication networks, telephone networks, wired and wireless television communication networks, etc. Examples of wireless data communication networks include 3G, 4G, 5G, 3GPP (3rd Generation Partnership Project), LTE (Long Term Evolution), WIMAX (World Interoperability for Microwave Access), Wi-Fi, Bluetooth communication, infrared communication, and ultrasound. This includes, but is not limited to, communication, Visible Light Communication (VLC), LiFi, etc.

The speaker authentication server 100 may acquire the real-time real-time speech quality of the speaker from the real-time counseling voice information of the speaker consulting with the counselor, and transmit the acquired real-time real-time speech quality of the speaker to the score calculation server 110. Here, the speaker's real-time real-time speech quality may be a voice in which the counselor's real-time real-time speech quality, silence period, and household noise are removed from the speaker's real-time counseling voice information. In addition, the speaker's real-time real-time speech loss may be a voice in which the speaker's simple answer real-time speech quality is further removed from the speaker's real-time consultation voice information.

The score calculation server 110 may perform real-time authentication of the speaker at each preset authentication cycle by comparing the real-time real-time voice of the speaker received from the speaker authentication server 100 with the speaker's pre-registered voice fingerprint.

The score calculation server 110 may calculate a real-time authentication score for the result of the speaker's real-time authentication and transmit the calculated real-time authentication score to the speaker authentication server 100.

The speaker authentication server 100 may determine the risk of impersonation for the speaker based on the real-time authentication score received from the score calculation server 110.

Hereinafter, the operation of each component of the speaker authentication system of FIG. 1 will be described in more detail.

FIG. 2 is a block diagram of the speaker authentication server 100 shown in FIG. 1 according to an embodiment of the present invention.

Referring to FIG. 2, the speaker authentication server 100 includes a voice fingerprint inquiry unit 200, a voice fingerprint registration unit 210, a real voice acquisition unit 220, a storage unit 230, a real voice transmission unit 240, It may include an authentication result receiving unit 250 and a risk determination unit 260. However, the speaker authentication server 100 shown in FIG. 2 is only one implementation example of the present invention, and various modifications are possible based on the components shown in FIG. 2.

The voice fingerprint search unit 200 may search the speaker's voice fingerprint through the score calculation server 110 based on the speaker's information (e.g., speaker identification information, such as the speaker's ID). Here, the score calculation server 110 can check whether the speaker's voice fingerprint is registered in the voice fingerprint storage (not shown) based on the speaker's information.

If the voice fingerprint of the speaker consulting with the counselor is not registered, the voice fingerprint registration unit 210 may obtain the stored actual voice quality of the speaker from the speaker's counseling voice information pre-stored in voice storage (not shown). For example, the voice fingerprint register 210 extracts the counselor's previously stored counseling voice information from the voice storage (not shown), and from the extracted speaker's counseling voice information, the counselor's stored real voice, silence period, household noise, and speaker It is possible to obtain the speaker's stored real speech by removing the simple answer real sound.

The voice fingerprint register 210 can determine whether the acquired stored real voice of the speaker is a voice confirmed as a normal consultation through analysis of a conversation between the speaker and a counselor and at the same time is a voice confirmed as the voice of a single speaker.

The storage unit 230 may map the acquired real voice of the speaker and the speaker information and store them in a learning storage (not shown).

The voice fingerprint registration unit 210 may transmit a voice fingerprint registration request including the acquired stored real voice of the speaker and speaker information (e.g., speaker identification information, speaker ID, etc.) to the score calculation server 110. .

In another embodiment, the voice fingerprint registration unit 210 provides real-time voice information of the speaker when the voice fingerprint of the speaker consulting with the counselor is not registered and the speaker's counseling voice information is not stored in voice storage (not shown). A voice fingerprint registration request including the speaker's real-time real-time speech obtained from and the speaker's information may be transmitted to the score calculation server 110.

At this time, the score calculation server 110, which has received the voice fingerprint registration request, learns the stored real voice or real-time real-time voice of the speaker included in the voice fingerprint registration request, and learns the speaker's information and the learned speaker's stored real voice or real-time voice. The speaker's voice fingerprint can be registered based on actual speech. For example, the score calculation server 110 may learn the speaker's stored real voice or real-time real voice to generate the speaker's voice feature vector, and register the generated speaker's voice feature vector as the speaker's voice fingerprint.

During counseling, the real-time speech acquisition unit 220 can acquire the real-time real-time speech quality of the speaker from the real-time consultation voice information of the speaker. For example, the real-time speech acquisition unit 220 removes the real-time real-time speech of the counselor, silent sections, and everyday noise (e.g., music, car engine noise, white noise, etc.) from the speaker's real-time consultation voice information, thereby removing the real-time real-time speech of the speaker. Voice can be obtained. Additionally, the real voice acquisition unit 220 may further remove the speaker's simple answer real voice (eg, 'yes, no', etc.) from the speaker's real-time consultation voice information.

The storage unit 230 may store the real-time real-time speech quality of the speaker along with the speaker information in a learning storage (not shown) using a sliding window method. Here, the sliding window method transmits data contained in the window size (a certain area of the memory buffer), and when reception of the data is confirmed, data is sequentially transmitted by moving the window step to the side (sliding). it means.

At this time, the speaker's real-time real-time speech quality may be stored in learning storage (not shown) from the start to the end of the speaker's consultation. For example, if the minimum real-time real-time voice quality for speaker authentication is secured in a voice shifted by a preset window step (e.g., within n seconds), it can be stored in the learning storage (not shown) using a sliding window method. You can.

The real speech transmission unit 240 may transmit the acquired real-time real speech of the speaker to the score calculation server 110. For example, referring to Figure 3A, the real speech transmission unit 240 may transmit the speaker's real-time real speech 30 acquired in real time during the consultation to the score calculation server 110 in real time. Here, the score calculation server 110, for example, calculates the real-time real-time voice of the speaker (30) and the speaker's pre-registered voice at each preset authentication cycle (e.g., 1 second cycle) until authentication of the speaker succeeds once. The speaker can be authenticated by comparing fingerprints.

The authentication result receiving unit 250 receives from the score calculation server 110 a real-time authentication score for the real-time authentication result of the speaker based on comparison between the speaker's real-time real-time voice at each preset authentication cycle and the speaker's pre-registered voice fingerprint. You can.

Here, the score calculation server 110 provides real-time information on the speaker's real-time authentication result calculated by comparing the speaker's real-time real-time voice and the speaker's pre-registered voice fingerprint, for example, after successful authentication of the speaker once. The authentication cycle can be varied based on the authentication score.

For example, referring to FIG. 3B, when the real-time authentication score is greater than the preset first score 301, the authentication cycle may be changed to the first authentication cycle 303 (eg, an 8 second cycle). Additionally, when the real-time authentication score is smaller than the preset first score 305, the authentication cycle may be changed to a second authentication cycle 303 (eg, a 4-second cycle). Here, the second authentication cycle 303 may be shorter than the first authentication cycle 303.

The risk determination unit 260 may determine the risk of impersonation for the speaker based on the real-time authentication score received from the score calculation server 110.

In addition, when determining the risk of impersonation for a speaker, the risk determination unit 260 uses the speaker's ID included in the speaker's information to request a customer center server (not shown) to inquire whether the speaker's age and gender match. , the risk of impersonation for the speaker can be determined based on whether the speaker's age and gender match received from the customer center server (not shown).

In addition, the risk determination unit 260 determines whether the speaker's voice fingerprint corresponds to a pre-stored risk group voice fingerprint, whether the speaker's age (or gender) matches the speaker, or whether the speaker may be separated during counseling (e.g., two people). The risk of impersonation for the speaker can be determined differently based on the case (e.g., when the above speakers take turns consulting).

Additionally, the risk determination unit 260 may provide information about the speaker's authentication result and impersonation risk to the counselor's terminal (not shown) during the consultation process with the speaker.

For example, referring to FIG. 3C, when the real-time authentication score falls within the first impersonation risk range, the risk determination unit 260 compares the pre-stored voice fingerprint of the risk group (criminal's voice fingerprint) with the speaker's voice fingerprint. The risk of impersonation of the speaker can be determined. If the speaker's voice fingerprint matches the pre-stored voice fingerprint of the risk group, the risk determination unit 260 determines the risk of impersonation for the speaker as high risk 309, which is the most dangerous level, and provides information about the determined risk of impersonation. and response manual information for the speaker can be provided to the counselor's terminal (not shown), while simultaneously transmitting an impersonation notification to the cyber police server (not shown). At this time, the risk determination unit 260 may conduct virtual counseling for a certain period of time to track the location of the criminal and secure the voice to help with the investigation of the criminal.

As another example, if the real-time authentication score falls within the second impersonation risk range, the risk determination unit 260 determines the impersonation risk for the speaker by determining whether the age range of the speaker included in the speaker information matches the speaker and the real-time authentication score. can be judged.

If the age range of the speaker does not match, the risk determination unit 260 determines the risk of impersonation for the speaker as medium risk (311) and sends information on the judged impersonation risk along with response manual information to the counselor's terminal (not shown). ) can be provided to.

As another example, when the authenticated speaker changes during counseling, the risk determination unit 260 transmits a speaker change notification message to the counselor's terminal (not shown) and guides confirmation of the relationship between the other changed speaker and the authenticated speaker. Response manual information may be provided to the counselor's terminal (not shown).

As another example, if the cause of the speaker's authentication failure is the voice quality of the speaker's real-time real-time speech, the risk determination unit 260 determines the risk of impersonation for the speaker as low risk (313) and Information on the risk of impersonation can be provided to the counselor's terminal (not shown).

Additionally, the risk determination unit 260 may determine whether authentication for the speaker is successful or fails based on the real-time authentication score. For example, if the speaker's authentication fails (i.e., impersonation) less than a preset number of times (1 to 2 times), the risk determination unit 260 provides a response manual to the counselor to induce the speaker to make the call. It can be provided through a terminal (not shown).

If the speaker's authentication fails more than a preset number of times, the impersonation risk can be set high, and information on the set impersonation risk as well as response manual information including termination of counseling can be provided to the counselor's terminal (not shown). . At this time, the actual voice of the speaker impersonating the speaker may be classified and stored as an impersonation voice in the learning storage (not shown).

The risk determination unit 260 may transmit information about the risk of impersonation of the speaker determined based on the real-time authentication score to the score calculation server 110. At this time, the score calculation server 110 may change the authentication cycle for comparison between the real-time real-time voice of the speaker and the speaker's pre-registered voice fingerprint based on the received information about the risk of impersonation of the speaker.

For example, referring to FIG. 3D, if the risk of impersonation of the speaker is high 315, the authentication cycle can be changed to the third authentication cycle 317 (eg, 1 second cycle). If the risk of impersonation of the speaker is medium risk (319), the authentication cycle may be changed to the fourth authentication cycle (312) (eg, 5 second cycle). If the risk of impersonation of the speaker is low (323), the authentication cycle may be changed to the fifth authentication cycle (325) (eg, a 10-second cycle).

Meanwhile, those skilled in the art will know that a voice fingerprint inquiry unit 200, a voice fingerprint registration unit 210, a real voice acquisition unit 220, a storage unit 230, a real voice transmission unit 240, an authentication result reception unit 250, and a risk level. It will be fully understood that each of the determination units 260 may be implemented separately, or one or more of them may be integrated and implemented.

Figure 4 is a flowchart showing a method of authenticating a speaker in the speaker authentication server 100, according to an embodiment of the present invention.

The speaker authentication method according to the embodiment shown in FIG. 4 includes steps processed in time series in the speaker authentication server 100 and the score calculation server 110 according to the embodiment shown in FIGS. 1 to 3D. Therefore, even if the content is omitted below, the content described regarding the speaker authentication server 100 and the score calculation server 110 of FIGS. 1 to 3D can also be applied to the speaker authentication method according to the embodiment shown in FIG. 4.

Referring to FIG. 4, in step S401, the speaker authentication server 100 may obtain the speaker's real-time real-time speech quality from the speaker's real-time consultation voice information.

In step S403, the speaker authentication server 100 may transmit the real-time real-time speech of the speaker to the score calculation server 110.

In step S405, the speaker authentication server 100 calculates a real-time authentication score for the speaker's real-time authentication result based on comparison between the speaker's real-time actual voice at each preset authentication cycle and the speaker's pre-registered voice fingerprint. It can be received from. Here, the authentication cycle can be varied based on the real-time authentication score.

In step S407, the speaker authentication server 100 may determine the risk of impersonation for the speaker based on the received authentication score.

In the above description, steps S401 to S407 may be further divided into additional steps or combined into fewer steps, depending on the implementation of the present invention. Additionally, some steps may be omitted or the order between steps may be changed as needed.

Figure 5 is a block diagram of the score calculation server 110 shown in Figure 1, according to an embodiment of the present invention.

Referring to FIG. 5, the score calculation server 110 includes a voice fingerprint registration unit 500, an actual voice reception unit 510, a real-time authentication performance unit 520, an authentication score calculation unit 530, and an authentication score transmission unit 540. and an authentication cycle setting unit 550. However, the score calculation server 110 shown in FIG. 5 is only one implementation example of the present invention, and various modifications are possible based on the components shown in FIG. 5.

The voice fingerprint registration unit 500 may receive from the speaker authentication server 100 a voice fingerprint registration request including the stored actual voice of the speaker and the speaker information obtained from the speaker's consultation voice information pre-stored in the voice storage (not shown). You can. Here, the speaker's stored real voice may be a voice in which the counselor's stored real voice, silence period, daily life noise, and the speaker's simple answer real voice are removed from the speaker's counseling voice information.

The voice fingerprint registration unit 500 may learn the stored real voice of the speaker included in the received voice fingerprint registration request, and register the speaker's voice fingerprint based on the speaker information and the learned stored real voice of the speaker. For example, the voice fingerprint registration unit 500 may learn the speaker's stored real voice, generate the speaker's voice feature vector, and register the generated speaker's voice feature vector as the speaker's voice fingerprint.

The real voice reception unit 510 can receive real-time real voice of a speaker who is consulting with a counselor from the speaker authentication server 100 in real time. For example, the real speech receiver 510 sends real-time real speech of the speaker, which includes only the voice of the speaker who is consulting with the counselor, and speaker information (e.g., speaker identification information, speaker ID, etc.) to the speaker authentication server ( 100). Here, the speaker's real-time real-time voice quality can be obtained from the speaker's real-time consultation voice information. For example, the speaker's real-time real-time speech is the voice in which the counselor's real-time real-time speech, silence sections, and household noise are removed from the speaker's real-time consultation voice information, and the speaker's simple answer real-time speech is further removed from the speaker's real-time consultation voice information. It could be a voice.

The real-time authentication unit 520 extracts a voice fingerprint that is mapped to the speaker's information among a plurality of voice fingerprints pre-registered in the voice fingerprint storage (not shown), and combines the extracted speaker's pre-registered voice fingerprint with the speaker's real-time actual voice fingerprint. By comparing voices, real-time authentication of the speaker can be performed at each preset authentication cycle. For example, the real-time authentication unit 520 compares the speaker's real-time real-time voice and the speaker's pre-registered voice fingerprint at each preset authentication cycle (e.g., 1 second cycle) until the speaker is successfully authenticated once. Speakers can be authenticated in real time.

The authentication score calculation unit 530 can calculate a real-time authentication score for the result of real-time authentication of the speaker. For example, the authentication score calculation unit 530 may calculate a real-time authentication score based on the similarity between the pre-registered voice fingerprint and the real-time real-time voice of the speaker.

The authentication score transmitter 540 may transmit the calculated real-time authentication score to the speaker authentication server 100.

The authentication cycle setting unit 550 can set the authentication cycle based on the real-time authentication score. For example, if the calculated real-time authentication score is greater than the preset first score, the authentication cycle setting unit 550 changes the authentication cycle to the first authentication cycle (e.g., 8 second cycle), and the real-time authentication score is If it is smaller than the first score, the authentication cycle can be changed to a second authentication cycle (eg, a 4-second cycle). Here, the second authentication cycle may be shorter than the first authentication cycle.

For example, after successful authentication of the speaker once, the authentication cycle setting unit 550 performs real-time authentication of the speaker's real-time authentication result calculated by comparing the speaker's real-time real-time voice and the speaker's pre-registered voice fingerprint. The certification cycle can be changed based on the score.

When the authentication cycle setting unit 550 receives information about the risk of impersonation for the speaker determined based on the real-time authentication score from the speaker authentication server 100, the authentication cycle setting unit 550 determines the risk of impersonation for the speaker based on the received information about the risk of impersonation for the speaker. The authentication cycle for comparison between the real-time real-time voice quality and the speaker's pre-registered voice fingerprint can be changed.

For example, if the risk of impersonating the speaker is high, the authentication cycle setting unit 550 may change the authentication cycle to a third authentication cycle (eg, 1 second cycle). Alternatively, if the risk of impersonating the speaker falls within the risk level, the authentication cycle setting unit 550 may change the authentication cycle to the fourth authentication cycle (eg, a 5-second cycle). Alternatively, if the risk of impersonation of the speaker falls under the risk level, the authentication cycle setting unit 550 may change the authentication cycle to the fifth authentication cycle (eg, a 10-second cycle).

Thereafter, the real-time authentication performing unit 520 may perform real-time authentication of the speaker by comparing the speaker's pre-registered voice fingerprint with the speaker's real-time real-time speech at each changed authentication cycle.

Thereafter, when the consultation between the counselor and the speaker is completed, the voice fingerprint registration unit 500 may update the speaker's voice fingerprint based on the real-time real voice that successfully authenticated the speaker and the speaker's pre-registered voice fingerprint. For example, the voice fingerprint registration unit 500 may generate a speaker's voice feature vector from real-time real-time speech that successfully authenticates the speaker, and re-register the generated speaker's voice feature vector as the speaker's voice fingerprint.

Meanwhile, those skilled in the art will know that the voice fingerprint registering unit 500, the real voice receiving unit 510, the real-time authentication performing unit 520, the authentication score calculating unit 530, the authentication score transmitting unit 540, and the authentication cycle setting unit 550. It will be fully understood that each may be implemented separately, or one or more of them may be implemented integratedly.

Figure 6 is a flowchart showing a method of authenticating a speaker in the score calculation server 110 according to an embodiment of the present invention.

The speaker authentication method according to the embodiment shown in FIG. 6 includes steps processed in time series in the speaker authentication server 100 and the score calculation server 110 according to the embodiment shown in FIGS. 1 to 5. Therefore, even if the content is omitted below, the content described regarding the speaker authentication server 100 and the score calculation server 110 of FIGS. 1 to 5 can also be applied to the speaker authentication method according to the embodiment shown in FIG. 6.

Referring to FIG. 6, in step S601, the score calculation server 110 may receive the real-time real-time speech quality of the speaker from the speaker authentication server 100. Here, the speaker's real-time real-time voice quality can be obtained from the speaker's real-time consultation voice information.

In step S603, the score calculation server 110 may perform real-time authentication of the speaker at each preset authentication cycle by comparing the real-time real-time voice of the speaker with the speaker's pre-registered voice fingerprint. Here, the authentication cycle can be varied based on the real-time authentication score.

In step S605, the score calculation server 110 may calculate a real-time authentication score for the result of the speaker's real-time authentication.

In step S607, the score calculation server 110 may transmit the real-time authentication score to the speaker authentication server 100.

In the above description, steps S601 to S607 may be further divided into additional steps or combined into fewer steps, depending on the implementation of the present invention. Additionally, some steps may be omitted or the order between steps may be changed as needed.

One embodiment of the present invention may also be implemented in the form of a recording medium containing instructions executable by a computer, such as program modules executed by a computer. Computer-readable media can be any available media that can be accessed by a computer and includes both volatile and non-volatile media, removable and non-removable media. Additionally, computer-readable media may include all computer storage media. Computer storage media includes both volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data.

The description of the present invention described above is for illustrative purposes, and those skilled in the art will understand that the present invention can be easily modified into other specific forms without changing the technical idea or essential features of the present invention. will be. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive. For example, each component described as unitary may be implemented in a distributed manner, and similarly, components described as distributed may also be implemented in a combined form.

The scope of the present invention is indicated by the claims described later rather than the detailed description, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. .

100: Speaker authentication server
110: Score calculation server
200: Voice fingerprint inquiry unit
210: Voice fingerprint register
220: Real voice acquisition unit
230: storage unit
240: Real voice transmission unit
250: Authentication result receiving unit
260: Risk judgment unit
500: Voice fingerprint register
510: Real voice receiver
520: Real-time authentication department
530: Certification score calculation unit
540: Certification score transmission unit
550: Authentication cycle setting unit

Claims (19)

In the speaker authentication server that authenticates the speaker consulting with the counselor,
a real-time speech acquisition unit that acquires the real-time real-time speech of the speaker from the real-time consultation voice information of the speaker;
a real-time speech transmission unit that transmits the real-time real-time speech of the speaker to a score calculation server;
an authentication result receiving unit that receives, from the score calculation server, a real-time authentication score for the real-time authentication result of the speaker based on comparison between the real-time real-time voice of the speaker at each preset authentication cycle and the speaker's pre-registered voice fingerprint; and
A risk determination unit that determines the risk of impersonation for the speaker based on the received real-time authentication score,
The authentication cycle is variable based on the real-time authentication score,
If the real-time authentication score is greater than the preset first score, the authentication cycle is changed to the first authentication cycle,
If the real-time authentication score is less than the first score, the authentication cycle is changed to a second authentication cycle,
A speaker authentication server, wherein the second authentication cycle is shorter than the first authentication cycle.
According to claim 1,
Obtaining the stored real voice of the speaker from the speaker's consultation voice information previously stored in voice storage, and transmitting a voice fingerprint registration request including the acquired stored real voice of the speaker and the speaker's information to the score calculation server. A speaker authentication server further comprising a voice fingerprint register.
According to claim 2,
A speaker authentication server further comprising a voice fingerprint search unit that searches the speaker's voice fingerprint through the score calculation server based on the speaker's information.
According to claim 1,
A speaker authentication server further comprising a storage unit that stores the acquired real-time real-time speech quality of the speaker in a learning storage using a sliding window method.
According to claim 1,
A speaker authentication server, wherein the real-time real-time speech of the speaker is a voice in which the real-time real-time speech of the counselor, silent sections, and household noise are removed from the real-time counseling voice information of the speaker.
According to claim 5,
A speaker authentication server, wherein the real-time real-time speech quality of the speaker is further removed from the speaker's real-time consultation voice information.
According to claim 5,
The score calculation server learns the received stored real voice of the speaker, and registers the speaker's voice fingerprint based on the speaker information and the learned stored real voice of the speaker.
delete In the speaker authentication server that authenticates the speaker consulting with the counselor,
a real-time speech acquisition unit that acquires the real-time real-time speech of the speaker from the real-time consultation voice information of the speaker;
a real-time speech transmission unit that transmits the real-time real-time speech of the speaker to a score calculation server;
an authentication result receiving unit that receives, from the score calculation server, a real-time authentication score for the real-time authentication result of the speaker based on comparison between the real-time real-time voice of the speaker at each preset authentication cycle and the speaker's pre-registered voice fingerprint; and
A risk determination unit that determines the risk of impersonation for the speaker based on the received real-time authentication score,
The authentication cycle is variable based on the real-time authentication score.
A speaker authentication server wherein the authentication cycle is variable based on the risk of impersonation for the speaker.
In the score calculation server that authenticates the speaker consulting with the counselor,
a real-time speech reception unit that receives real-time real-time speech of the speaker from a speaker authentication server;
a real-time authentication unit that performs real-time authentication of the speaker at preset authentication cycles by comparing the speaker's real-time real-time voice quality with the speaker's pre-registered voice fingerprint;
an authentication score calculation unit that calculates a real-time authentication score for the result of real-time authentication of the speaker; and
Authentication score transmission unit that transmits the real-time authentication score to the speaker authentication server
Including,
The real-time real-time speech quality of the speaker is obtained from the real-time consultation voice information of the speaker,
The authentication cycle is variable based on the real-time authentication score,
The authentication cycle setting unit
If the real-time authentication score is greater than the preset first score, change the authentication cycle to the first authentication cycle,
If the real-time authentication score is less than the first score, change the authentication cycle to a second authentication cycle,
The score calculation server, wherein the second authentication cycle is shorter than the first authentication cycle.
According to claim 10,
It further comprises a voice fingerprint registration unit that receives a voice fingerprint registration request including the stored actual voice of the speaker and the speaker information obtained from the speaker's consultation voice information pre-stored in voice storage from the speaker authentication server, Score calculation server.
According to claim 11,
The voice fingerprint registration unit learns the stored real voice of the speaker, and registers the speaker's voice fingerprint based on the speaker's information and the learned stored real voice of the speaker.
According to claim 11,
The voiceprint registering unit updates the speaker's voiceprint based on the real-time actual voice and the speaker's pre-registered voiceprint after the consultation between the counselor and the speaker is terminated.
In the score calculation server that authenticates the speaker consulting with the counselor,
a real-time speech reception unit that receives real-time real-time speech of the speaker from a speaker authentication server;
a real-time authentication unit that performs real-time authentication of the speaker at preset authentication cycles by comparing the speaker's real-time real-time voice quality with the speaker's pre-registered voice fingerprint;
an authentication score calculation unit that calculates a real-time authentication score for the result of real-time authentication of the speaker; and
Authentication score transmission unit that transmits the real-time authentication score to the speaker authentication server
Including,
The real-time real-time speech quality of the speaker is obtained from the real-time consultation voice information of the speaker,
The authentication cycle is variable based on the real-time authentication score,
A score calculation server further comprising an authentication cycle setting unit that sets the authentication cycle based on the real-time authentication score.
delete In a method of authenticating a speaker consulting with a counselor in a speaker authentication server,
Obtaining the real-time real-time speech quality of the speaker from the real-time consultation voice information of the speaker;
Transmitting the real-time real-time speech of the speaker to a score calculation server;
Receiving a real-time authentication score for the real-time authentication result of the speaker from the score calculation server based on comparison between the real-time real-time voice of the speaker at each preset authentication cycle and the speaker's pre-registered voice fingerprint; and
A step of determining a risk of impersonation for the speaker based on the received real-time authentication score,
The authentication cycle is variable based on the real-time authentication score,
If the real-time authentication score is greater than the preset first score, the authentication cycle is changed to the first authentication cycle,
If the real-time authentication score is less than the first score, the authentication cycle is changed to a second authentication cycle,
A speaker authentication method, wherein the second authentication cycle is shorter than the first authentication cycle.
According to claim 16,
Obtaining the stored real voice of the speaker from the speaker's consultation voice information previously stored in voice storage, and transmitting a voice fingerprint registration request including the acquired stored real voice of the speaker and the speaker's information to the score calculation server. A speaker authentication method further comprising steps.
According to claim 17,
A speaker authentication method further comprising querying the speaker's voice fingerprint through the score calculation server based on the speaker information.
According to claim 16,
A speaker authentication method further comprising the step of storing the acquired real-time real-time speech quality of the speaker in a learning storage using a sliding window method.

Images