Voice processing system, program, method, online conference system, and voice communication device

Abstract

To provide a technique for continuously authenticating a person by voice.SOLUTION: A voice processing system has: a storage unit that stores voiceprint data representing voice characteristics of a specific person; a reception processing unit that acquires the voice of a person contained in continuously received voice data; and an authentication processing unit that performs voice print authentication on the acquired voice in chronological order on the basis of the voice print data to generate authentication result values in chronological order, which indicate the degree to which the voice is estimated to be that of the specified person.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present disclosure relates to techniques for performing voice authentication. [Background technology]

[0002] Voice data is used for various purposes. Voice data is used to communicate over the telephone or in online conferences. Voice data is also used to create meeting minutes. In these various applications, there is a need to identify the voice of a specific person from the sounds contained in the data.

[0003] There is a voice authentication technology that verifies the identity of a person from their voice. Voice authentication is used to verify the identity of a caller at a call center. It is also used to unlock a smartphone (see Patent Document 1). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Special Publication No. 2015-509680 Summary of the Invention [Problem to be solved by the invention]

[0005] Generally, voice authentication technology is used for identity verification. Therefore, once identity verification is complete, voice authentication is also completed. However, there are situations where it is required to continuously identify a person. For example, there are cases where it is necessary to extract only the voice of a specific person from the voice of an online conference or telephone call. Also, in order to prevent various special frauds using telephones, there are cases where it is necessary to continuously measure the voice of a telephone call to see if the other party is a malicious person.

[0006] One object of the present disclosure is to provide a technology for continuously authenticating a person using voice. [Means for solving the problem]

[0007] A voice processing system according to one aspect of the present disclosure includes a memory unit that stores voiceprint data representing characteristics of the voice of a specific person, a reception processing unit that acquires the person's voice contained in continuously received voice data, and an authentication processing unit that performs voiceprint authentication on the acquired voice in a time series based on the voiceprint data, and generates authentication result values ​​in a time series that represent the degree to which the voice is estimated to be that of the specific person.

[0008] An online conference system according to another aspect of the present disclosure is an online conference system for multiple participants to communicate with each other by voice via a communication network, and includes a voice processing system that processes voice data collected from the participants and generates voice data to be heard by the other participants, the voice processing system including a memory unit that stores voiceprint data representing characteristics of the voice of a specific person, a reception processing unit that acquires the voice of the person contained in continuously received voice data, and a voice processing unit that performs voiceprint authentication on the acquired voice in chronological order based on the voiceprint data, and outputs an authentication result that indicates the degree to which the voice is estimated to be the voice of the specific person. the authentication result processing unit is capable of selecting a pass mode for passing the voice of the specific person and a discard mode for discarding the voice of the specific person, and in the pass mode, passes the voice whose authentication result value is equal to or greater than the threshold, and passes the voice whose authentication result value is below the threshold if it satisfies a predetermined condition, and in the discard mode, discards the voice whose authentication result value is equal to or greater than the threshold, and passes the voice whose authentication result value is below the threshold.

[0009] According to another aspect of the present disclosure, a voice communication device is a voice communication device for a user to communicate with a call partner via a communication network, the voice processing system processing voice data collected from the call partner's voice and generating voice data to be heard by the user, the voice processing system having a memory unit storing voiceprint data representing the characteristics of the voice of a specific person, a reception processing unit acquiring the person's voice contained in continuously received voice data, an authentication processing unit performing voiceprint authentication on the acquired voice in a time series based on the voiceprint data and generating authentication result values ​​in a time series representing the degree to which the voice is estimated to be the voice of the specific person, and an authentication result processing unit executing processing depending on whether the authentication result value is equal to or greater than a predetermined threshold, the memory unit further storing context data in which words are registered, the authentication result processing unit determining whether a voice whose authentication result value is below the threshold is suspicious based on the context data, and if suspicious, executing one or more of recording the call, sending a warning email, and issuing a voice warning. Effect of the Invention

[0010] According to one aspect of the present disclosure, it becomes possible to perform continuous person authentication using voice. [Brief description of the drawings]

[0011] [Figure 1] 1 is a schematic block diagram of an online conference system according to a first embodiment. [Diagram 2] 1 is a block diagram of a voice processing system according to a first embodiment. [Diagram 3] 4 is a flowchart of a process executed by the voice processing system according to the first embodiment. [Figure 4] 3 is a table showing system common data in the first embodiment. [Diagram 5] 4 is a table showing individual data in the first embodiment. [Figure 6] 13 is a table showing another example of individual data in the first embodiment. [Figure 7]FIG. 11 is a block diagram of a voice processing system according to a modified example of the first embodiment. [Figure 8] FIG. 11 is a schematic block diagram of a voice communication device according to a second embodiment. [Figure 9] FIG. 11 is a block diagram of a voice processing system according to a second embodiment. [Figure 10] 11 is a flowchart of a process executed by the voice processing system according to the second embodiment. [Figure 11] 13 is a table showing system common data in the second embodiment. [Figure 12] 13 is a table showing individual data in the second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. EXAMPLES

[0013] In the first embodiment, an online conference system capable of allowing the other party to hear only the voice of a participant from among voices picked up by a microphone will be exemplified.

[0014] FIG. 1 is a schematic block diagram of an online conference system according to a first embodiment.

[0015] Referring to FIG. 1, the online conference system 11 is a system that realizes a dialogue between multiple participants using voice or voice and image. The online conference system 11 incorporates a voice processing system 21. The voice processing system 21 is a system that identifies and extracts the voices of participants 911 and 912 from collected voices. Note that FIG. 1 shows an example in which two participants 911 and 912 are at different locations and hold a one-to-one conference using the online conference system 11, but this is not limited to this form. There may be three or more locations where the participants 91 are present. In addition, there may be multiple participants 91 at one location. In addition, the number of participants 91 at one location whose voice is identified and extracted is not limited to one. The voices of multiple participants 91 at one location may be identified and extracted.

[0016] As an example, the online conference system 11 is configured with a center system (not shown) constructed on a cloud and terminal devices (not shown) such as personal computers, smartphones, or tablet terminals used by the participants 911 and 912. In this case, the voice processing system 21 may be incorporated in the system on the cloud or in the terminal devices.

[0017] Fig. 2 is a block diagram of the voice processing system in Example 1. The voice processing system 21 in Fig. 2 is implemented on a cloud as an example here, receives voice data (received voice data) from terminal devices (not shown) of participants 911 and 912 via the Internet 92, extracts the voices of the participants 911 and 912 from the received voice data, generates transmission voice data from the extracted voices, and transmits the transmission voice data to the terminal devices (not shown) of the participants 912 and 911 via the Internet 92.

[0018] Here, as an example, a process of extracting the voice of participant 911 from received voice data from the terminal device of participant 911, generating transmission voice data from the extracted voice, and transmitting it to the terminal device of participant 912 is assumed to be the same process as a process of extracting the voice of participant 912 from received voice data from the terminal device of participant 912, generating transmission voice data from the extracted voice, and transmitting it to the terminal device of participant 911. The following description will focus mainly on the process of extracting the voice of participant 911 from received voice data from the terminal device of participant 911, generating transmission voice data from the extracted voice, and transmitting it to the terminal device of participant 912.

[0019] The voice processing system 21 includes a central processing unit 31 and a memory unit 32 .

[0020] The central processing unit 31 includes a CPU (not shown) and a memory (not shown), and the system is realized by executing a software program. The central processing unit 31 realizes a communication unit 33, a voice receiving unit 34, a voice transmitting unit 35, an authentication management system 36, and an authentication result determination system 40. The authentication management system 36 includes a voice information reception processing unit 37, a voice information authentication processing unit 38, and an authentication result output unit 39. The authentication result determination system 40 includes an authentication result processing unit 41.

[0021] The storage unit 32 is a storage device that stores data used for processing the software program. The storage unit 32 stores a voiceprint database (voiceprint DB) 42 in which voiceprint data representing the voice characteristics of the participants 911 and 912 are registered.

[0022] The communication unit 33 is a communication device that realizes data communication via the Internet 92 .

[0023] The voice receiving unit 34 receives the received voice data from the terminal device of the participant 911 and provides it to the authentication management system 36 .

[0024] The authentication management system 36 continuously performs voiceprint authentication on the received voice data in chronological order, and outputs in chronological order the authentication result of the participant 911. The authentication result of the participant 911 is, for example, an index value (authentication result value) indicating the degree to which the voice is estimated to be that of the participant 911.

[0025] In the authentication management system 36, the voice information reception processing unit 37 receives the received voice data from the voice receiving unit 34 and acquires the voice of the person. Here, the voices of multiple people may be acquired, for example, the voice of the participant 911 and the voice of another person may be acquired. The voice information authentication processing unit 38 performs voiceprint authentication on the person's voice acquired from the received voice data in time series based on the voiceprint data, and generates authentication result values ​​in time series. The authentication result value indicates the degree to which the voice is estimated to be that of the participant 911, so the larger the value, the more likely it is that the voice is that of the participant 911. The authentication result output unit 39 outputs the person's voice acquired from the received voice data and the authentication result value of the voice as the authentication result.

[0026] The authentication result determination system 40 executes processing on the received voice data based on the authentication result by the authentication management system 36 .

[0027] In the authentication result judgment system 40, the authentication result processing unit 41 sequentially compares the time-series authentication result values ​​output by the authentication result output unit 39 of the authentication management system 36 with a predetermined threshold, and while the authentication result value is equal to or greater than the threshold, determines that the voice of the person corresponding to that authentication result value is the voice of the participant 911, and generates audio data containing only the voice of the participant 911.

[0028] The voice transmitting unit 35 transmits the voice data containing only the voice of the participant 911 generated by the authentication result processing unit 41 as transmission voice data to the terminal device of the participant 912 via the communication unit 33 and the Internet 92. As a result, in the terminal device of the participant 912, only the voice of the participant 911 included in the voice data collected by the terminal device of the participant 911 is output and heard by the participant 912.

[0029] The processing of the voice processing system 21 will now be described in further detail.

[0030] 3 is a flowchart of the process executed by the voice processing system in the first embodiment. The voice processing system 21 executes the process using system common data and individual data. The system common data is common data that does not depend on an individual person. When the whole or a part of the voice processing system 21 is packaged as a general-purpose system, the overall operation mode may be set as the system common data. The individual data is data recorded for each individual person.

[0031] FIG. 4 is a table showing system common data in the first embodiment.

[0032] The system-wide data includes settings for the following items: voiceprint authentication threshold, email sending threshold, action mode, voice adjustment function, context check, voice recording, sending email address, and email sending / warning to caller.

[0033] The voiceprint authentication threshold is a threshold value that is compared with the authentication result value to determine whether the voice of a person acquired from voice data is that of a person registered in the voiceprint DB 42. If the authentication result value is equal to or greater than this threshold, it is determined that the person whose voice is acquired from the voice data is the same person as the person whose voiceprint data is registered in the voiceprint DB 42. In this embodiment, the voiceprint authentication threshold is set to 80%.

[0034] The email sending threshold is a threshold that is compared with the authentication result value in order to send a predetermined email according to the authentication result value. If the authentication result value falls below this threshold, a predetermined email is sent. For example, it can be used to automatically send an email if the person whose voice is acquired from the voice data is not the person whose voiceprint data is registered in the voiceprint DB 42. In this embodiment, the function of sending email is not used, so the email sending threshold is not set.

[0035] The action mode is setting information that determines the action to be taken when the authentication result value is equal to or greater than the voiceprint authentication threshold. When the authentication result value is equal to or greater than the voiceprint authentication threshold, it is possible to set whether to allow the person's voice to pass, to discard the voice, or to do nothing even if the authentication result value is equal to or greater than the voiceprint authentication threshold.

[0036] The voice adjustment execution function is setting information that specifies the operation of discard processing when the authentication result value falls below the voiceprint authentication threshold. The discard processing here is not simply a process of discarding the voice, but different processing is executed depending on the set mode. This discard processing is called the voice adjustment execution function. The voice adjustment execution function has three modes: automatic, manual, and off. When the automatic mode is set, for the voice of a person whose authentication result value is below the voiceprint authentication threshold, if the number of people whose voice is acquired from the voice data is less than a predetermined number, the voice is passed, and when the number exceeds the predetermined number, only the voice of the predetermined number of people whose authentication result value is a high value and whose volume is the loudest is passed. Note that this predetermined number may be set by the user. For example, when one person participates in an online conference, it may be set to one person. When the manual mode is set, the user can select whether to pass or discard the voice of a person whose authentication result value is below the voiceprint authentication threshold. When the off mode is set, the voice of a person whose authentication result value is below the voiceprint authentication threshold is discarded. In this embodiment, the voice adjustment execution function is set to automatic.

[0037] The context check is setting information that specifies whether or not to check the context of the voice when the authentication result value falls below the voiceprint authentication threshold. If the context check is set to on, the context of the voice whose authentication result value falls below the voiceprint authentication threshold is checked. The context check is, for example, a check as to whether or not a preset word appears in the spoken utterance. For example, the context check can be used to register words that frequently appear in suspicious utterances in advance, and determine that the person making the voice is highly likely to be suspicious if the word appears.

[0038] Voice recording is setting information that specifies whether or not to record the voice when the authentication result value falls below the voiceprint authentication threshold but the context check does not determine the voice as suspicious. If voice recording is set to on, the voice is recorded even when the authentication result value falls below the voiceprint authentication threshold but the context check does not determine the voice as suspicious. If voice recording is set to off, the voice is not recorded even when the authentication result value falls below the voiceprint authentication threshold but the context check does not determine the voice as suspicious. In this embodiment, voice recording is set to off.

[0039] The sending email address is setting information that specifies the destination address to which an email is sent when the authentication result value falls below the email sending threshold. The sending email / warning to caller is setting information that specifies whether or not to send an email or output a warning when the authentication result value falls below the email sending threshold. In this embodiment, sending an email or outputting a warning is not set, and therefore the sending email address is not set either.

[0040] FIG. 5 is a table showing individual data in the first embodiment. The individual data records voiceprint information, name, and statistical information for each person. The voiceprint information is voiceprint information that indicates the voice characteristics of the person. Multiple pieces of voiceprint information can be set for one person. The name is the person's name or nickname. The statistical information records statistical information obtained from the usage history of the person. The statistical information records a conference counter, conference time, and conference history. The conference counter is the number of times the person participated in a conference. The conference time is the time the person participated in a conference. The conference history is the date and time of the conference in which the person participated.

[0041] Fig. 6 is a table showing another example of individual data in the embodiment 1. In the example of Fig. 6, unlike the examples shown in Fig. 4 and Fig. 5, the voiceprint authentication threshold can be set for each person.

[0042] 3, first, in step 101, the voice processing system 21 acquires a person's voice from the received voice data by the voice information reception processing unit 37. Then, in step 102, the voice processing system 21 performs voiceprint authentication on the acquired voice based on the voiceprint data registered in the voiceprint DB 42 by the voice information authentication processing unit 38. As a result, an authentication result value is calculated for the acquired voice.

[0043] In step 103, the voice processing system 21 determines whether the action mode is set to the pass mode by the voice information authentication processing unit 38. If the action mode is set to the pass mode, in step 104, the voice processing system 21 determines whether the authentication result value is equal to or greater than the voiceprint authentication threshold by the voice information authentication processing unit 38. If the authentication result value is equal to or greater than the voiceprint authentication threshold, in step 105, the voice processing system 21 updates the statistical information of the voice and the person whose authentication result value is equal to or greater than the voiceprint authentication threshold by the authentication result output unit 39, displays on the screen that the voice is to be passed, and passes the voice by the authentication result processing unit 41. The voice passed by the authentication result processing unit 41 is transmitted from the voice transmission unit 35 to the terminal devices of the other participants in the conference via the communication unit 33 and the Internet 92, which is an example of a communication network.

[0044] If the authentication result value is below the voiceprint authentication threshold in step 104, the voice processing system 21 determines in step 106 whether the voice adjustment implementation function is set to ON by the voice information authentication processing unit 38. Here, ON has two modes, automatic and manual. If the voice adjustment implementation function is set to manual, the voice processing system 21, in step 107, asks the user (conference participant) via the voice information authentication processing unit 38 whether to pass the voice whose authentication result value is below the voiceprint authentication threshold. At this time, the voice information authentication processing unit 38 displays the person whose voiceprint has been authenticated, the authentication result value and threshold, and information about the voice (e.g., a text display of the voice, a judgment button), and prompts the user to input a judgment. On the other hand, if the voice adjustment implementation function is set to automatic, the voice processing system 21 skips step 107 and proceeds to step 108.

[0045] In step 108, the voice processing system 21 uses the voice information authentication processing unit 38 to determine whether or not to pass the voice. At this time, if the voice adjustment execution function is set to automatic, the determination is made by a predetermined automatic determination, and if the voice adjustment execution function is set to manual, the determination is made according to the determination input by the user in step 107. The automatic determination determines whether or not to pass the voice depending on the automatic mode of the discarding process described above. That is, for voices of persons whose authentication result value is below the voiceprint authentication threshold, if there is only one person whose voice has been acquired from the voice data, the voice is passed, and if there are multiple people, only the voice with the highest authentication result value and the loudest volume is passed.

[0046] If it is determined that the voice should be passed, the voice processing system 21 registers the voice as a voice to be passed by the voice information authentication processing unit 38 in step 109. Furthermore, in step 110, the voice processing system 21 updates the statistical information of the voice by the voice information authentication processing unit 38, displays on the screen that the voice should be passed by the authentication result output unit 39, and passes the voice of the person by the authentication result processing unit 41.

[0047] If it is determined in step 108 that the voice should not be allowed to pass, the voice processing system 21 in step 112 causes the voice information authentication processing unit 38 to update the statistical information for the voice, causes the authentication result output unit 39 to display on the screen that the voice should not be allowed to pass, and causes the authentication result processing unit 41 to discard the voice.

[0048] If the voice adjustment implementation function is set to off in step 106, the voice processing system 21 in step 111 updates the statistical information of the voice whose authentication result value is below the voiceprint authentication threshold by the voice information authentication processing unit 38, displays on the screen a message that the voice will be discarded by the authentication result output unit 39, and discards the voice by the authentication result processing unit 41.

[0049] If the action mode is not set to pass in step 103, the voice processing system 21 determines in step 113 whether the action mode is set to discard by the voice information authentication processing unit 38. If the action mode is set to discard, the voice processing system 21 determines in step 114 whether the authentication result value is equal to or greater than the voiceprint authentication threshold by the voice information authentication processing unit 38. If the authentication result value is equal to or greater than the voiceprint authentication threshold, the voice processing system 21 updates the statistical information of the voice and the person whose authentication result value is equal to or greater than the voiceprint authentication threshold by the voice information authentication processing unit 38 in step 115, displays on the screen a message to the effect that the voice is to be discarded by the authentication result output unit 39, and discards the voice by the authentication result processing unit 41. If the authentication result value is below the voiceprint authentication threshold, the voice processing system 21 updates the statistical information of the voice whose authentication result value is below the voiceprint authentication threshold by the authentication result output unit 39 in step 116, displays on the screen a message to the effect that the voice is to be passed, and passes the voice by the authentication result processing unit 41.

[0050] The online conference system 11 of the present embodiment may display various information on the management screen based on data set or acquired by the voice processing system 21. For example, the management screen may display setting information on whether the function of passing and discarding voice by voiceprint authentication is on or off, setting information on whether to pass or discard voice whose authentication result value is equal to or greater than the voiceprint authentication threshold, the voiceprint authentication threshold, setting information on the voice adjustment implementation function, and further the number of people setting, etc. Also, for example, the number of people whose voice is acquired from the received voice data, the number of people whose voice is being passed, information on the people whose voice is being passed, the number of people whose voice is being discarded, the authentication result value of the voice being passed, a light indicating that the voice is being passed, etc. may be displayed on the online conference screen. Also, a function for the participant 911 to check in advance whether his / her voice can be passed may be provided, and a test button for activating the function may be displayed on the management screen or the online conference screen. Also, a warning sound may be used to notify that his / her voice has been acquired from the received voice data but discarded by voiceprint authentication.

[0051] According to the embodiment described above, the voice processing system 21 includes a storage unit 32 that stores voiceprint data representing the characteristics of a specific person's voice, a voice information reception processing unit 37 that acquires the person's voice contained in continuously received voice data, and a voice information authentication processing unit 38 that performs voiceprint authentication on the acquired voice in time series based on the voiceprint data and generates authentication result values ​​in time series that represent the degree to which the voice is estimated to be that of a specific person. As a result, voiceprint authentication is performed on the received voice data in time series based on the voiceprint data representing the characteristics of the voice of the specific person, and authentication result values ​​in time series that represent the degree to which the voice is estimated to be that of a specific person are generated, making it possible to continuously authenticate a person by voice.

[0052] The voice processing system 21 further includes an authentication result processing unit that compares the authentication result value with a predetermined threshold, and generates voice data of only the voice of the specific person in the received voice data while the authentication result value is equal to or greater than the threshold. Therefore, since voice data of the voice is generated while a condition is satisfied by the voiceprint authentication, only voices that satisfy the condition can be passed through the voiceprint authentication.

[0053] According to the present embodiment, the online conference system 11 is an online conference system for a plurality of participants 911, 912 to have a conversation with each other by voice via a communication network 95, and includes a voice processing system 21 that processes first voice data obtained by collecting the voice of the first participant 911 and generates second voice data to be heard by the second participant 912. The voice processing system 21 includes a storage unit 32 that stores voiceprint data representing characteristics of the voice of a specific person, a voice information reception processing unit 37 that acquires the voice of the person contained in continuously received voice data, and a voice information reception processing unit 37 that performs voiceprint authentication on the acquired voice in chronological order based on the voiceprint data, and detects when the voice is estimated to be the voice of the specific person. and an authentication result processing unit 41 which compares the authentication result value with a predetermined voiceprint authentication threshold and processes the received voice data depending on whether the authentication result value is equal to or greater than the voiceprint authentication threshold. The authentication result processing unit 41 can select a pass mode which passes the voice of a specific person and a discard mode which discards the voice of a specific person. In the pass mode, it passes voice whose authentication result value is equal to or greater than the voiceprint authentication threshold and passes voice whose authentication result value is below the voiceprint authentication threshold if it satisfies a predetermined condition. In the discard mode, it discards voice whose authentication result value is equal to or greater than the voiceprint authentication threshold and passes voice whose authentication result value is below the voiceprint authentication threshold if it satisfies a predetermined condition.

[0054] As one example, the predetermined condition is that the user permits passage. As another example, the predetermined condition is that when the number of people whose voices are acquired from the received voice data is equal to or less than a predetermined number, the voices of the people are allowed to pass, and when the number of people whose voices are acquired from the received voice data is greater than the predetermined number, only the voices of the predetermined number of people selected based on the authentication result value and the volume are allowed to pass.

[0055] In the above embodiment, the voice processing system 21 is implemented on the cloud, but it may be implemented on a terminal device such as a personal computer, a smartphone, or a tablet terminal used by the participants 911 and 912.

[0056] FIG. 7 is a block diagram of a voice processing system in a modified example of the first embodiment. Referring to FIG. 7, the voice processing system 21 implemented on the terminal device is different from the voice processing system 21 shown in FIG. 2 in that it does not include a communication unit 33. The voice of the participant 911 is collected by a microphone 51 provided in or connected to a terminal device (not shown), and the received voice data is received by a voice receiving unit 34 of the voice processing system 21. In addition, the transmitted voice data from the voice transmitting unit 35 is transmitted to the Internet 92 via a communication unit 52 provided in the terminal device. Here, an example is shown in which the voice processing system 21 implemented on the terminal device uses the communication unit 52 provided in the terminal device for communication, but other configurations are also possible. For example, a communication unit may be provided in the voice processing system 21 as in the voice processing system 21 on the cloud shown in FIG. 2. EXAMPLES

[0057] In the second embodiment, a voice communication device that continuously measures from a call voice whether a call partner is a malicious person or not is exemplified.

[0058] FIG. 8 is a schematic block diagram of a voice communication device according to the second embodiment.

[0059] 8, the voice communication device 12 is a terminal device that realizes voice communication, such as a smartphone, a mobile phone, a landline phone, or a personal computer. A user 93 can use the voice communication device 12 to make a voice call with a call partner via a communication network such as the Internet or a mobile communication network (not shown). A voice processing system 21 is built into the voice communication device 12. The voice processing system 21 is a system that identifies a preset call partner 94 and performs a predetermined process for a call with another call partner.

[0060] Fig. 9 is a block diagram of a voice processing system in Example 2. Voice processing system 21 in Fig. 9 differs from voice processing system 21 shown in Fig. 7 in that an e-mail system 44 and an e-mail transmission unit 46 are realized by central processing unit 31, and a context database (context DB) 43 is stored in storage unit 32. E-mail system 44 includes an e-mail processing unit 45. Voice processing system 21 in Fig. 9 also receives incoming voice data via communication network 95 and communication unit 52. The transmitted voice data from voice transmission unit 35 is converted into voice by speaker 53, and is heard by user 93.

[0061] The authentication result processing unit 41 executes processing according to the authentication result using the authentication result value from the authentication result output unit 39 as well as the context data registered in the context DB 43. The context data is data for determining the context of a suspicious call. A list of words that appear in a suspicious call is registered in the context data. The email system 44 is a system equipped with a function for sending email. In response to an instruction from the authentication result processing unit 41, the email processing unit 45 generates a preset email addressed to a preset destination. The generated email is sent by the email sending unit 46 to a predetermined destination via the communication unit 52 and the communication network 95.

[0062] FIG. 10 is a flowchart of a process executed by the voice processing system according to the second embodiment.

[0063] Fig. 11 is a table showing system common data in Example 2. The system common data in Example 2 is similar to that in Example 1 shown in Fig. 4. However, in Example 2, the context check is set to OFF, voice recording is set to ON, the sending email address is set, and sending email / warning to caller is set to ON.

[0064] FIG. 12 is a table showing individual data in Example 2. The items recorded as statistical information in the individual data in Example 2 are different from those in Example 1 shown in FIG. 5. The statistical information in Example 2 is recorded as statistical information acquired from a call history with a person registered in advance. The statistical information in Example 2 records a call counter, a call time, and a call history. The call counter is the number of times a call was made with the person. The call time is the time when a call was made with the person. The call history is the date and time when a call was made with the person.

[0065] 10, first, in step 101, the voice processing system 21 acquires a person's voice from the received voice data by the voice information reception processing unit 37. Furthermore, in step 102, the voice processing system 21 performs voiceprint authentication on the acquired voice based on the voiceprint data registered in the voiceprint DB 42 by the voice information authentication processing unit 38. As a result, an authentication result value is calculated for the acquired voice.

[0066] In step 201, the voice processing system 21 determines whether or not the authentication result value is equal to or greater than the voiceprint authentication threshold using the voice information authentication processing unit 38. If the authentication result value is equal to or greater than the voiceprint authentication threshold, in step 202, the voice processing system 21 updates the statistical information of the voice and the person whose authentication result value is equal to or greater than the voiceprint authentication threshold using the authentication result output unit 39, and allows the voice to pass through using the authentication result processing unit 41, allowing the normal call to continue.

[0067] If the authentication result value falls below the voiceprint authentication threshold in step 201, the voice processing system 21 determines whether or not to perform a context check using the voice information authentication processing unit 38 in step 203. Whether or not to perform a context check can be determined by whether the context check in the system common data shown in Fig. 11 is on or off.

[0068] If it is determined that the context check is not to be performed, in step 204, the voice processing system 21 updates the statistical information by the authentication result output unit 39, starts recording the call voice by the authentication result processing unit 41, alerts the user 93, and causes the email processing unit 45 to generate an email for a warning. The email processing unit 45 generates an email with preset contents addressed to a destination email address registered in advance. The sending email address in the system common data shown in FIG. 11 is used as the destination email address. The generated email is sent by the email sending unit 46 to the destination email address via the communication unit 52 and the communication network 95. The alert is issued, for example, by a screen display or audio output.

[0069] If it is determined in step 203 that a context check is to be performed, in step 205, the speech processing system 21 checks the context of the speech in question using the authentication result processing unit 41. The context check is a check as to whether or not the speech contains words that frequently appear in suspicious statements and that are preset in the context DB 43. If a word registered in the speech in question appears, the context of the speech is presumed to be suspicious.

[0070] In step 206, the voice processing system 21 determines whether the context of the voice is suspicious or not through the authentication result processing unit 41. If the context of the voice is suspicious, in step 207, the voice processing system 21 updates the statistical information through the authentication result output unit 39, starts recording the call voice through the authentication result processing unit 41, alerts the user 93, and causes the email processing unit 45 to generate a warning email. The email processing unit 45 generates an email with preset contents addressed to a destination email address registered in advance.

[0071] If the context is determined not to be suspicious in step 206, the voice processing system 21 updates the statistical information by the authentication result output unit 39 in step 208, passes the voice, and continues the normal call. At this time, if voice recording is set to ON in the system common data shown in Fig. 11, the voice processing system 21 starts recording the voice of the call by the authentication result processing unit 41.

[0072] In this embodiment, the voice processing system 21 performs voiceprint authentication continuously while the call is continuing, but other operations are possible. For example, voiceprint authentication may be performed continuously from the start of the call until a predetermined duration has elapsed, and voiceprint authentication may not be performed thereafter. For example, voiceprint authentication may be performed until 30 seconds have elapsed from the start of the call, and then terminated.

[0073] The voice communication device 12 of the present embodiment may display various information on the screen based on data set or acquired by the voice processing system 21. For example, setting information on whether the voiceprint authentication processing function is on or off, a voiceprint authentication threshold, a destination email address to which a warning email is sent, words registered in the context DB 43, etc. may be displayed on the management screen. Also, for example, information on a person registered in the voiceprint DB 42 matching the voice of the call, a message that a person registered in the voiceprint DB 42 matching the voice of the call cannot be found, etc. may be displayed. Also, the voice communication device 12 may generate an alarm sound based on data set or acquired by the voice processing system 21. For example, an alarm sound may be generated when a person registered in the voiceprint DB 42 matching the voice of the call cannot be found.

[0074] According to the embodiment described above, the storage unit 32 stores context data indicating a predetermined context, and the authentication result processing unit 41 compares the authentication result value with the voiceprint authentication threshold, and if the authentication result value is continuously below the voiceprint authentication threshold, extracts the voice context in the received voice data, and performs different processing depending on whether the extracted context matches the context indicated in the context data stored in the storage unit 32. In this way, when the authentication result value falls below the threshold, the context is further inspected, and different processing is performed depending on the inspection result, so that a call from a suspicious person can be detected by using voiceprint authentication and context inspection in combination. For example, it is possible to detect a special fraud call in which the context from a suspicious person pretending to be a relative contains malicious voice, and perform some processing.

[0075] Furthermore, the email processing unit 45 has a function of generating a predetermined email addressed to a predetermined destination address, and when the extracted context matches the context indicated in the context data, the authentication result processing unit 41 causes the email processing unit 45 to generate the email. This makes it possible to notify a desired destination by email that a call suspected to be a malicious call from a suspicious person has arrived, for example.

[0076] Furthermore, the authentication result processing unit 41 compares the authentication result value with the voiceprint authentication threshold, and if the authentication result value is continuously below the voiceprint authentication threshold, causes the email processing unit 45 to generate an email.

[0077] Furthermore, the voice information authentication processing unit 38 performs voiceprint authentication on the acquired voice for a predetermined duration, and ends voiceprint authentication on the voice when the duration has elapsed. In this way, by performing continuous voiceprint authentication for a predetermined duration, it is possible to efficiently detect suspicious calls.

[0078] The voice communication device 12 of this embodiment is a voice communication device for a user 93 to communicate with a call partner 94 via a communication network 95, and includes a voice processing system 21 that processes first voice data obtained by collecting the voice of the call partner 94 and generates second voice data for the user 93 to hear. The voice processing system 21 includes a memory unit 32 that stores voiceprint data representing characteristics of a specific person's voice, a voice information reception processing unit 37 that acquires the person's voice contained in continuously received voice data, and a voice information reception processing unit 38 that performs voiceprint recognition in chronological order for the acquired voice based on the voiceprint data. the voice information authentication processing unit 38 which performs voice authentication to generate, in a time series, authentication result values ​​which indicate the degree to which the voice is estimated to be that of a specific person, and an authentication result processing unit 41 which compares the authentication result value with a voiceprint authentication threshold and executes processing according to whether the authentication result value is equal to or greater than the voiceprint authentication threshold, the memory unit 32 further stores context data in which words are registered, and the authentication result processing unit 41 determines whether a voice whose authentication result value is below the voiceprint authentication threshold is suspicious based on the context data, and if it is suspicious, executes one or more of recording the call, sending a warning e-mail, and issuing a voice warning.

[0079] In addition, in the present embodiment, an example has been shown in which the voice processing system 21 is implemented in the voice call device 12, which is a terminal device, but other configurations are also possible. The voice processing system 21 may be implemented anywhere on the path along which voice data is transmitted in a voice call. For example, a system for relaying voice data may be provided on the cloud, and the voice processing system 21 may be incorporated into that system.

[0080] The above-described embodiments of the present invention are merely illustrative examples of the present invention, and are not intended to limit the scope of the present invention to these embodiments. Those skilled in the art can implement the present invention in various other forms without departing from the scope of the present invention. [Explanation of symbols]

[0081] 11...online conference system, 12...voice communication device, 21...voice processing system, 31...central processing unit, 32...storage unit, 33...communication unit, 34...voice receiving unit, 35...voice transmitting unit, 36...authentication management system, 37...voice information receiving processing unit, 38...voice information authentication processing unit, 39...authentication result output unit, 40...authentication result judgment system, 41...authentication result processing unit, 42...voiceprint database, 43...context database, 44...email system, 45...email processing unit, 46...email transmitting unit, 51...microphone, 52...communication unit, 53...speaker, 92...internet, 93...user, 94...call partner, 95...communication network, 911...participant, 912...participant

Claims

1. A storage unit for storing voiceprint data representing characteristics of a voice of a specific person; A reception processing unit for acquiring a person's voice included in continuously received voice data; an authentication processing unit that performs voiceprint authentication on the acquired voice in a time series based on the voiceprint data and generates an authentication result value in a time series that indicates a degree to which the voice is estimated to be that of a specific person; A voice processing system having:

2. an authentication result processing unit that generates voice data including only the voice of the specific person in the received voice data or voice data excluding the voice of the specific person while the authentication result value is equal to or greater than a predetermined threshold value; The audio processing system of claim 1 .

3. The storage unit further stores context data indicating a predetermined context, and an authentication result processing unit that extracts a voice context from the received voice data if the authentication result value is below a predetermined threshold, and performs different processes depending on whether the extracted context matches a context indicated in the context data. The audio processing system of claim 1 .

4. further comprising an email processing unit for generating a predetermined email addressed to a predetermined destination address; the authentication result processing unit causes the electronic mail processing unit to generate the electronic mail if the extracted context matches a context indicated in the context data; The audio processing system of claim 3.

5. an email processing unit that generates a predetermined email addressed to a predetermined destination address; and an authentication result processing unit that causes the email processing unit to generate the email if the authentication result value is below a predetermined threshold. The audio processing system of claim 1 .

6. the authentication processing unit performs voiceprint authentication on the acquired voice for a predetermined duration, and terminates the voiceprint authentication on the voice when the duration has elapsed.

6. A voice processing system according to claim 3 or 5.

7. the authentication result processing unit is capable of selecting a pass mode for generating voice data including only the voice of the specific person in the received voice data while the authentication result value is equal to or greater than a predetermined threshold value, and a discard mode for generating voice data from which the voice of the specific person has been removed. The audio processing system of claim 2 .

8. On the computer, Voiceprint data that represents the characteristics of a specific person's voice is stored, Receive continuous incoming voice data; A voice processing program for executing the steps of: performing voiceprint authentication on the received voice data in a time series manner based on the voiceprint data; and generating authentication result values ​​in a time series manner which indicate the degree to which the voice data is estimated to be that of a specific person.

9. The computer Voiceprint data that represents the characteristics of a specific person's voice is stored, Receive continuous incoming voice data; a voice processing method for executing voiceprint authentication on the received voice data in a time series manner based on the voiceprint data, and generating authentication result values ​​in a time series manner which indicate the degree to which the voice is estimated to be that of the specific person.

10. An online conference system for enabling a plurality of participants to have a voice conversation with each other via a communication network, comprising: a voice processing system that processes voice data collected from the participants and generates voice data to be heard by other participants; The voice processing system includes: A storage unit for storing voiceprint data representing characteristics of a voice of a specific person; A reception processing unit for acquiring a person's voice included in continuously received voice data; an authentication processing unit that performs voiceprint authentication on the acquired voice in a time series based on the voiceprint data and generates an authentication result value in a time series that indicates a degree to which the voice is estimated to be that of a specific person; an authentication result processing unit that processes the received voice data according to whether the authentication result value is equal to or greater than a predetermined threshold value, the authentication result processing unit is capable of selecting a pass mode for passing the voice of the specific person and a discard mode for discarding the voice of the specific person, In the pass mode, a voice whose authentication result value is equal to or greater than the threshold value is passed, and a voice whose authentication result value is less than the threshold value is passed if a predetermined condition is satisfied. In the discard mode, a voice whose authentication result value is equal to or greater than the threshold value is discarded, and a voice whose authentication result value is less than the threshold value is passed. Online conference system.

11. The predetermined condition is that the user has given permission to pass through. The online conference system according to claim 10.

12. The predetermined condition is that, when the number of people whose voices are acquired from the received voice data is equal to or less than a predetermined number, the voices of the people are passed, and, when the number of people whose voices are acquired from the received voice data is greater than the predetermined number, only the voices of the predetermined number of people selected based on the authentication result value and the volume are passed. The online conference system according to claim 10.

13. A voice communication device for enabling a user to make a voice call with a call partner via a communication network, comprising: a voice processing system that processes first voice data obtained by collecting the voice of the call partner and generates second voice data to be heard by the user; The voice processing system includes: A storage unit for storing voiceprint data representing characteristics of a voice of a specific person; A reception processing unit for acquiring a person's voice included in continuously received voice data; an authentication processing unit that performs voiceprint authentication on the acquired voice in a time series based on the voiceprint data and generates an authentication result value in a time series that indicates a degree to which the voice is estimated to be that of a specific person; an authentication result processing unit that executes a process according to whether the authentication result value is equal to or greater than a predetermined threshold value, The storage unit further stores context data in which the words are registered, The authentication result processing unit: For a voice whose authentication result value is below the threshold, determining whether or not the voice is suspicious based on the context data, and if the voice is suspicious, executing one or more of recording the call, sending a warning e-mail, and issuing a voice warning. Voice communication device.

Images