Voiceprint processing method and device, equipment, medium and program product

Abstract

The invention discloses a voiceprint processing method and device, equipment, a medium and a program product, which are used for improving the identity authentication accuracy based on voiceprint characteristics in a call process. The voiceprint processing method comprises the following steps: extracting a first voiceprint feature of a first call voice of a speaker; querying a second voiceprint feature of the speaker in a voiceprint library; and based on a first similarity between the first voiceprint feature and the second voiceprint feature, and an information source of the first voiceprint feature and an information source of the second voiceprint feature, fusing the first voiceprint feature and the second voiceprint feature to obtain a third voiceprint feature of the speaker.

Description

Technical Field

[0001] The present application relates to the field of speech processing technology, and in particular to a voiceprint processing method, device, equipment, medium and program product. Background Art

[0002] As one of the main methods of biometric identification, voiceprint recognition mainly authenticates the speaker's voice. Due to its convenience and stability, it is gradually being applied in various fields. For example, voice is an effective way for speakers to communicate with each other, and it is also one of the effective ways to authenticate the speaker's identity. During a call, the speaker's identity authentication based on voiceprint features is gaining more and more attention and recognition.

[0003] How to improve the accuracy of identity authentication based on voiceprint features during calls has become a hot research topic. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a voiceprint processing method, device, equipment, medium and program product for improving the accuracy of identity authentication based on voiceprint features during a call.

[0005] In order to achieve the above objectives, the present application embodiment adopts the following technical solutions:

[0006] In a first aspect, an embodiment of the present application provides a voiceprint processing method, including:

[0007] Extracting a first voiceprint feature of a first call speech of a speaker;

[0008] Querying a voiceprint database for a second voiceprint feature of the speaker;

[0009] Based on the first similarity between the first voiceprint feature and the second voiceprint feature, the information source of the first voiceprint feature and the information source of the second voiceprint feature, the first voiceprint feature and the second voiceprint feature are fused to obtain a third voiceprint feature of the speaker.

[0010] In a second aspect, an embodiment of the present application provides a voiceprint processing device, including:

[0011] An acquisition module, used for extracting a first voiceprint feature of a first call speech of a speaker;

[0012] A query module, used for querying the second voiceprint feature of the speaker in the voiceprint library;

[0013] An updating module is used to fuse the first voiceprint feature and the second voiceprint feature based on a first similarity between the first voiceprint feature and the second voiceprint feature, an information source of the first voiceprint feature, and an information source of the second voiceprint feature to obtain a third voiceprint feature of the speaker.

[0014] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0015] processor;

[0016] a memory for storing instructions executable by the processor;

[0017] The processor is configured to execute the instructions to implement the voiceprint processing method provided in the first aspect.

[0018] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium. When instructions in the storage medium are executed by a processor of an electronic device, the electronic device can perform the voiceprint processing method provided in the first aspect.

[0019] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute part or all of the steps in the voiceprint processing method provided in the first aspect.

[0020] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects:

[0021] Extract the first voiceprint feature of the speaker's first call speech, and query the speaker's second voiceprint feature from the voiceprint library; the similarity between the speaker's second voiceprint feature and the first voiceprint feature can reflect whether the first voiceprint feature and the second voiceprint feature are the same person. On this basis, based on the information source of the first voiceprint feature and the information source of the second voiceprint feature, the first voiceprint feature and the second voiceprint feature are fused to obtain the speaker's third voiceprint feature, which not only makes the information source of the third voiceprint feature include the information source of the first voiceprint feature, overcoming the cross-information source comparison problem in the identity authentication process, but also effectively eliminates the impact of voiceprint changes on the accuracy of identity authentication, thereby improving the accuracy of identity authentication. In addition, the above method can be used to realize real-time and automatic update of the second voiceprint feature during a call, without the need for the speaker to register repeatedly, reducing resource consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0023] Figure 1 A schematic diagram of an implementation environment applicable to the voiceprint processing method provided in an embodiment of the present application;

[0024] Figure 2 A flow chart of a voiceprint processing method provided for one embodiment of the present application;

[0025] Figure 3 A schematic diagram of a login process of a call center system provided for one embodiment of the present application;

[0026] Figure 4 A schematic diagram of a voiceprint authentication process provided for an embodiment of the present application;

[0027] Figure 5 A schematic diagram of a voiceprint update process provided for an embodiment of the present application;

[0028] Figure 6 A schematic diagram of the structure of a voiceprint processing device provided in one embodiment of the present application;

[0029] Figure 7 A schematic diagram of the structure of an electronic device provided for one embodiment of the present application. DETAILED DESCRIPTION

[0030] To make the purpose, technical scheme and advantages of the present application clearer, the technical scheme of the present application will be clearly and completely described below in conjunction with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application. The terms "first", "second", etc. in this specification and claims are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here. In addition, "and / or" in this specification and claims represents at least one of the connected objects, and the character " / " generally represents that the front and back associated objects are an "or" relationship.

[0031] The applicant has studied the voiceprint-based identity authentication methods in related technologies and found that such methods usually require the speaker to pre-register the voiceprint features and then compare the registered voiceprint features of the speaker with the voiceprint features of the conversation to achieve the identity authentication of the speaker. In this process, the reasons for the low accuracy of identity authentication are mainly the following two aspects:

[0032] On the one hand, the voice used by the speaker to register the voiceprint features is usually a computer recording, while the call voice during the call is usually collected by a telephone recording device. When the communication operator transmits the call voice, in order to ensure the continuity of the call content, the call voice will be compressed and encoded, which leads to the situation that the registration voice and the call voice cross the information source. Therefore, identity authentication based on the voiceprint features of the registration voice against the voiceprint features of the call voice will reduce the authentication accuracy due to the cross-information source.

[0033] On the other hand, voiceprint recognition is time-varying, that is, the speaker's voiceprint characteristics will change over time, and thus the voiceprint recognition performance will decrease over time, affecting the accuracy of identity authentication.

[0034] In view of this, the embodiment of the present application proposes a voiceprint processing method, which extracts the first voiceprint feature of the speaker's first call speech, and queries the speaker's second voiceprint feature from the voiceprint library; the similarity between the speaker's second voiceprint feature and the first voiceprint feature can reflect whether the first voiceprint feature and the second voiceprint feature are the same person. On this basis, based on the information source of the first voiceprint feature and the information source of the second voiceprint feature, the first voiceprint feature and the second voiceprint feature are fused to obtain the speaker's third voiceprint feature, which not only makes the information source of the third voiceprint feature include the information source of the first voiceprint feature, overcoming the cross-information source comparison problem in the identity authentication process, but also effectively eliminates the impact of voiceprint changes on the accuracy of identity authentication, thereby improving the accuracy of identity authentication. In addition, the above method can be used to realize real-time and automatic update of the second voiceprint feature during a call, without the need for the speaker to register repeatedly, reducing resource consumption.

[0035] It should be understood that the voiceprint processing method provided in the embodiment of the present application can be executed by an electronic device. The electronic device here can include terminal devices, such as smart phones, tablet computers, laptops, desktop computers, intelligent voice interaction devices, smart home appliances, smart watches, vehicle terminals, aircraft, etc.; or, the electronic device can also include a server, such as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0036] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0037] In order to facilitate the understanding of the technical solution provided in the embodiments of the present application, firstly Figure 1 , the implementation environment applicable to the technical solution provided in the embodiment of the present application is described. It should be understood that the technical solution provided in the embodiment of the present application is applicable to Figure 1The implementation environment shown is only an exemplary description and should not be understood as a limitation on the implementation environment to which the technical solution is applicable.

[0038] like Figure 1 As shown, an implementation environment applicable to the technical solution provided in the embodiment of the present application includes a terminal device 1 and a server 2. The terminal device 1 interacts with the server 2 via a wireless network or a wired network. The terminal device may include, for example, but is not limited to, a smart phone, a tablet computer, a laptop computer, a desktop computer, an intelligent voice interaction device, a smart home appliance, a smart watch, a vehicle terminal, an aircraft, etc. The server 2 may be an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. A voiceprint library 21 is deployed in the server 2.

[0039] As an example, an application supporting voice collection function is running in the terminal device 1. In the voiceprint registration stage, the terminal device 1 can collect the voice of the speaker, extract the voiceprint features from the voice and send them to the server 2, which will be added to the voiceprint library 3 as the voiceprint features registered by the speaker for identity authentication.

[0040] In order to prevent others from impersonating the speaker to make a call, during the call, server 2 extracts voiceprint features from the speaker's current call voice, and compares the extracted voiceprint features with the speaker's registered voiceprint features for similarity, and determines whether the speaker has passed the identity authentication based on the comparison results.

[0041] Considering that the voiceprint features registered by the speaker and the voiceprint features of the conversation voice may come from different information sources, and the voiceprint features of the speaker change over time, cross-information source comparison and voiceprint feature changes will affect the accuracy of identity authentication. To this end, the server 2 can also update the registered voiceprint features based on the voiceprint features of the speaker's conversation voice to overcome the above problems.

[0042] Based on the above implementation environment, this application embodiment proposes a voiceprint processing method. Figure 2 , is a flow chart of a voiceprint processing method provided by an embodiment of the present application, the method comprising the following steps:

[0043] S202: Extract a first voiceprint feature of a first call speech of a speaker.

[0044] The speaker can be any party in the call process. For example, in an outbound call scenario, in order to prevent the agent from asking someone else to replace him / her to communicate with the customer, it is often necessary to authenticate the agent based on the voiceprint feature. For this purpose, the agent can be used as the speaker.

[0045] In the above S202, the first voiceprint feature of the speaker may be obtained in various ways, which may be selected according to actual needs, and the embodiment of the present application does not limit this.

[0046] In one embodiment, the first call speech is taken as a whole to perform voiceprint extraction to obtain the first voiceprint feature. This method has the advantages of simple implementation and high extraction efficiency.

[0047] In another embodiment, multiple voice segments are obtained from the first call voice; the voiceprint features of each voice segment are extracted; and the voiceprint features of the multiple voice segments are averaged to obtain the first voiceprint feature.

[0048] The average here can be an equal-weighted average or a weighted average.

[0049] As an example, multiple voice segments are obtained from the first call voice; the voiceprint features of each voice segment are extracted; and the voiceprint features of the multiple voice segments are equally weighted averaged to obtain the first voiceprint feature.

[0050] For example, by performing voice detection on the first call voice and removing the silent part, three valid voice segments are obtained, including voice segments 1 to 3. Then, voiceprint feature e1 is extracted from voice segment 1, voiceprint feature e2 is extracted from voice segment 2, and voiceprint feature e3 is extracted from voice segment 3. Further, the three voiceprint features are averaged with equal weights to obtain the first voiceprint feature E, that is, E = (e1 + e2 + e3) / 3.

[0051] The first voiceprint feature obtained by the above method can more accurately reflect the characteristics of the speaker and help reduce the impact of voiceprint changes on the identity authentication result. In practical applications, when the recognition error rate decreases slightly with the increase in time, that is, the voiceprint recognition has good robustness, the first voiceprint feature can be determined in this way.

[0052] As another example, multiple voice segments are obtained from the first call voice; the voiceprint feature of each voice segment is extracted; and based on the length of each voice segment, the multiple voice segments are weighted averaged to obtain the first voiceprint feature.

[0053] For example, still taking the above three speech segments and their voiceprint features as an example, the duration of each speech segment is used as the weight of the speech segment, and the voiceprint features of the three speech segments are weighted averaged to obtain the first voiceprint feature E, that is, E = (t1*e1+t2*e2+t3*e3) / (t1+t2+t3), t1 represents the duration of speech segment 1, t2 represents the duration of speech segment 2, and t3 represents the duration of speech segment 3.

[0054] Since the voiceprint features of longer voice segments are more reliable, the voiceprint features of all voice segments are weighted averaged based on the duration of each voice segment. The first voiceprint feature obtained can be more accurate and can more effectively reduce the impact of voiceprint changes on identity authentication results. In practical applications, when the recognition error rate decreases significantly with the increase in duration, that is, the voiceprint recognition robustness is poor, it is more appropriate to use this method to determine the first voiceprint feature.

[0055] S204: Search the voiceprint database for the second voiceprint feature of the speaker.

[0056] The voiceprint database stores multiple second voiceprint features. The second voiceprint feature refers to the voiceprint feature that is pre-registered in the voiceprint database and serves as the identity authentication standard for the corresponding speaker. The second voiceprint feature here can be the voiceprint feature registered by the speaker using computer recording before the first call, or the voiceprint feature updated during the call. The second voiceprint feature and the first voiceprint feature usually come from different information sources, and the voiceprint feature updated during the call integrates the first voiceprint feature, so its information source is the same as the information source of the first voiceprint feature.

[0057] It is worth mentioning that the information source of the voiceprint feature refers to the source of the voiceprint feature, such as from computer recording or from call voice.

[0058] S206: Based on the first similarity between the first voiceprint feature and the second voiceprint feature, the information source of the first voiceprint feature and the information source of the second voiceprint feature, the first voiceprint feature and the second voiceprint feature are merged to obtain a third voiceprint feature of the speaker.

[0059] After the third voiceprint feature is obtained, the third voiceprint feature can be added to the voiceprint library to replace the second voiceprint feature and used as a basis for authenticating the speaker.

[0060] The similarity between the first voiceprint feature and the second voiceprint feature can reflect whether the first voiceprint feature and the second voiceprint feature are of the same person. On this basis, based on the information source of the first voiceprint feature and the information source of the second voiceprint feature, the second voiceprint feature is updated to obtain the third voiceprint feature, which not only makes the information source of the third voiceprint feature the same as that of the first voiceprint feature, that is, both are derived from the call voice, thus overcoming the cross-information source comparison problem in the identity authentication process, but also effectively eliminates the impact of voiceprint changes on the accuracy of identity authentication, thereby improving the accuracy of identity authentication. In addition, the above method also realizes the real-time and automatic update of the second voiceprint feature during the call, without the need for the speaker to register repeatedly, reducing resource consumption.

[0061] In order to achieve real-time and automatic update of the second voiceprint feature to better utilize the speaker's physical state changes and voiceprint changes, the above S206 includes the following steps:

[0062] S262: Determine whether to update the second voiceprint feature based on the first similarity, the information source of the first voiceprint feature, and the information source of the second voiceprint feature.

[0063] As an example, if the information source of the first voiceprint feature is different from the information source of the second voiceprint feature, and the first similarity is greater than or equal to a first preset similarity threshold, it is determined to update the second voiceprint feature.

[0064] Exemplarily, during the first call, since the second voiceprint feature of the speaker has not been updated, the information source of the queried second voiceprint feature is the recording used when registering through the recording device, which is different from the information source of the first voiceprint feature (i.e., the call voice). In the case where the first similarity between this type of second voiceprint feature and the first voiceprint feature is greater than or equal to the first preset similarity threshold, it can be determined that the two voiceprint features belong to the same speaker. In this case, it is determined to update the second voiceprint feature so that the first voiceprint feature can be used to update the second voiceprint feature later, so that the information source of the updated third voiceprint feature is the same as the information source of the first voiceprint feature, that is, both are call voice. In this way, when the third voiceprint feature is used to authenticate the speaker, the problem of cross-information source comparison in the authentication process can be effectively overcome, thereby improving the accuracy of authentication.

[0065] As another example, if the information source of the first voiceprint feature is the same as the information source of the second voiceprint feature, and the first similarity is less than a first preset similarity threshold, the first identity feature of the speaker is obtained; the speaker is authenticated based on the first identity feature; if the speaker passes the authentication, it is determined to update the second voiceprint feature.

[0066] Exemplarily, after the second voiceprint feature in the voiceprint library has been updated, the information source of the second voiceprint feature queried from the voiceprint library is the same as the information source of the first voiceprint feature, that is, both are call voices. If the first similarity between such second voiceprint feature and the first voiceprint feature is less than the first preset similarity threshold, it may be because the two voiceprint features belong to different speakers, or because the speaker's voiceprint feature during the call has changed due to a change in physical state (such as a sudden cold). To this end, the speaker's first identity feature is used for identity authentication. If the authentication is successful, it is determined that the first voiceprint feature and the second voiceprint feature belong to the same person. In this case, it is determined to update the second voiceprint feature so that the first voiceprint feature can be used to update the second voiceprint feature later, so that the third voiceprint feature can adapt to the change in the speaker's physical state. In this way, when the third voiceprint feature is used to authenticate the speaker, the influence of the voiceprint change on the accuracy of identity authentication is effectively eliminated, thereby improving the accuracy of identity authentication.

[0067] In practical applications, the first identity feature can be selected according to actual needs, and the embodiments of the present application do not limit this. For example, the first identity feature can include facial features, so that the speaker is authenticated by means of face recognition. For another example, the first identity feature can also include fingerprint features, so that the speaker is authenticated by means of fingerprint recognition.

[0068] As another example, if the information source of the first voiceprint feature is the same as the information source of the second voiceprint feature, and the first similarity is greater than or equal to the first preset similarity threshold, then the duration between the generation time point of the second voiceprint feature and the determination time point of the first similarity is determined; if the duration is less than the first preset duration, and the first similarity is less than the second preset similarity threshold, then it is determined to update the second voiceprint feature; if the duration is greater than the second preset duration, or the duration is not less than the first preset duration and the first similarity is less than the second preset similarity threshold, then it is determined to update the second voiceprint feature. Wherein, the second preset duration is greater than the first preset duration.

[0069] For the second voiceprint feature of the original registration, its generation time point refers to the time point when the speaker is registered in the voiceprint database. For example, if the speaker registers the second voiceprint feature in the voiceprint database on January 1, 2024, the generation time point of the second voiceprint feature is January 1, 2024. For the second voiceprint feature of the non-original registration, its generation time point refers to the time when it was last updated. For example, if the second voiceprint feature of the original registration is updated on February 1, 2024, and voiceprint feature 1 is obtained, the generation time point of voiceprint feature 1 is February 1, 2024.

[0070] The time point for determining the first similarity refers to the time point for performing a similarity comparison between the first voiceprint feature of the first call speech and the second voiceprint feature.

[0071] The second preset similarity threshold can be set according to actual needs, and the embodiment of the present application does not limit this. For example, under normal circumstances, the similarity of voiceprint comparison is 0.8, and the first preset similarity threshold is 0.7, then the second preset similarity threshold is a value between the two, such as 0.75.

[0072] Both the first preset duration and the second preset duration can be determined according to the voiceprint similarity over different time spans. For example, if the voiceprint similarity over different time spans does not change much, the voiceprint recognition has a stronger ability to resist time variations, and the first preset duration and the second preset duration can be set to higher values; if the voiceprint similarity over different time spans changes greatly, the voiceprint recognition has a weaker ability to resist time variations, and the first preset duration and the second preset duration can be set to lower values.

[0073] Exemplarily, if the information source of the first voiceprint feature is the same as the information source of the second voiceprint feature, and the first similarity between the second voiceprint feature and the first voiceprint feature is greater than or equal to the first preset similarity threshold, it can be determined that the two voiceprint features belong to the same speaker. In this case, the duration between the generation time point of the first voiceprint feature and the determination time point of the similarity and the first similarity are combined to determine whether to update the second voiceprint feature, so as to ensure that the updated third voiceprint feature can adapt to the changes in the speaker's physical state. In this way, when the third voiceprint feature is used to authenticate the speaker, the influence of voiceprint changes on the accuracy of identity authentication is effectively eliminated, thereby improving the accuracy of identity authentication.

[0074] When the duration is less than the first preset duration, it is the initial stage of generating the second voiceprint feature, and the similarity between the first voiceprint feature and the second voiceprint feature is usually high. If the first similarity is less than the second preset similarity threshold at this stage, it is determined that the speaker's voiceprint feature has changed significantly at this time, and then the second voiceprint feature is updated, so that the first voiceprint feature can be used to update the second voiceprint feature later, so that the updated third voiceprint feature can adapt to the changes in the speaker's physical state. In this way, when the third voiceprint feature is used to authenticate the speaker, the influence of the voiceprint change on the accuracy of identity authentication is effectively eliminated, thereby improving the accuracy of identity authentication.

[0075] When the duration is not less than the first preset duration, it belongs to the middle and late stages of the generation of the second voiceprint feature. At this stage, the voiceprint feature of the same speaker will change significantly compared to the second voiceprint feature, and the similarity between the two will decrease as the duration increases. Based on this, if the duration is greater than the second preset duration, it is determined to update the second voiceprint feature, which can ensure that the second voiceprint feature is updated in a timely manner, and then when performing identity authentication based on the updated third voiceprint feature, the adverse effects of the voiceprint change can be effectively eliminated, thereby improving the accuracy of identity authentication.

[0076] If the duration is not less than the first preset duration and the first similarity is less than the second preset similarity threshold, it is determined to update the second voiceprint feature. This can prevent the second voiceprint feature from deviating from the actual situation due to the speaker's health condition (such as a sudden cold), improve the reliability and real-time performance of the second voiceprint feature, and then, when identity authentication is performed based on the second voiceprint feature, it can effectively eliminate the adverse effects of voiceprint changes and the deviation of the second voiceprint feature from the actual situation, thereby improving the accuracy of identity authentication.

[0077] S264: If the second voiceprint feature is updated, the first voiceprint feature and the second voiceprint feature are merged to obtain a third voiceprint feature of the speaker.

[0078] The updating of the obtained voiceprint feature based on the first voiceprint feature can be implemented in various appropriate ways, which is not limited in the embodiments of the present application.

[0079] As an example, the first voiceprint feature is used as the third voiceprint feature.

[0080] As another example, the above S262 includes the following steps: Step A1, determining a first candidate voiceprint feature of the speaker based on the first voiceprint feature; Step A2, fusing the first candidate voiceprint feature and the second voiceprint feature to obtain a third voiceprint feature of the speaker.

[0081] In this way, not only can it be ensured that the information source of the updated third voiceprint feature is the same as the information source of the first voiceprint feature to overcome the problem of cross-information source comparison, but it can also ensure that the third voiceprint feature is applicable to changes in the speaker's physical state while ensuring the reliability and accuracy of the updated third voiceprint feature, so as to effectively eliminate the adverse effects brought about by voiceprint changes, thereby improving the accuracy of identity authentication.

[0082] In step A1, in one method, the first voiceprint feature can be used as the first candidate voiceprint feature of the speaker. This method has the advantages of being convenient and fast.

[0083] In another manner, the voiceprint feature of the second call speech of the speaker is obtained, and the voiceprint feature of the second call speech and the first voiceprint feature are averaged to obtain the first candidate voiceprint feature of the speaker.

[0084] The second call voice refers to the call voice generated before the first call voice.

[0085] Considering that determining the first candidate voiceprint feature of the speaker based only on the first voiceprint feature may lead to homogeneity of the third voiceprint feature obtained based on it, for example, they are all voiceprint features when the speaker is communicating calmly with the other party, which affects the accuracy of identity authentication. Therefore, since the second call speech of the speaker contains the voiceprint features of the speaker in different states, by averaging the first voiceprint feature of the speaker and the voiceprint features of the second call speech, the homogeneity problem of the third voiceprint feature can be avoided, and the accuracy of identity authentication can be further improved.

[0086] In the embodiment of the present application, the second call voice may include the call voice of the speaker within M (M is an integer greater than 1) days before the first call voice.

[0087] For example, the voiceprint features of all second call speech and the first voiceprint feature may be averaged, so that the first candidate voiceprint feature obtained contains richer information, covering the voiceprint features of the speaker in various states.

[0088] Exemplarily, a third call voice can also be selected from the second call voice, wherein the second similarity between the voiceprint feature of the third call voice and the second voiceprint feature is greater than or equal to a third preset similarity threshold; the voiceprint feature of the third call voice and the first voiceprint feature are averaged to obtain the first candidate voiceprint feature of the speaker.

[0089] For example, the speaker has a total of 5 call voices in the past M days, including call voice 1 to call voice 5. The similarity between the voiceprint feature of each call voice and the second voiceprint feature is calculated respectively. If the similarity between the voiceprint features of call voice 1 and call voice 2 and the second voiceprint feature is greater than or equal to the third preset similarity threshold, the voiceprint feature of call voice 1, the voiceprint feature of call voice 2 and the first voiceprint feature are averaged to obtain the first candidate voiceprint feature of the speaker, which is used to update the second voiceprint feature.

[0090] It can be understood that the second voiceprint feature with a higher second similarity to the first voiceprint feature has a higher reliability and can accurately reflect the real voiceprint feature of the speaker. By averaging this type of voiceprint feature with the first voiceprint feature, the reliability of the first candidate voiceprint feature can be ensured, thereby ensuring the reliability of the third voiceprint feature obtained based on this update, thereby improving the accuracy of identity authentication.

[0091] Exemplarily, a candidate call voice set can also be generated based on the first call voice and the second call voice; a fourth call voice is selected from the candidate call voice set, wherein the third similarity between the voiceprint features of the fourth call voice and the second voiceprint features is greater than or equal to a fourth preset similarity threshold; based on the voiceprint features of the fourth call voice, a fifth call voice is selected from the candidate call voice set, wherein the fourth similarity between the voiceprint features of the fifth call voice and the second voiceprint features is less than the fourth preset similarity threshold, and the fifth similarity between the voiceprint features of the fifth call voice and the fourth call voice is greater than or equal to the fifth preset similarity threshold; the voiceprint features of the fourth call voice and the voiceprint features of the fifth call voice are averaged to obtain the first candidate voiceprint features of the speaker.

[0092] For example, the speaker has a total of 5 call voices in the past M days, including call voice 1 to call voice 5, then the candidate call voice set includes the first call voice and all call voices 1 to call voice 5. The similarity between the voiceprint feature of each call voice in the candidate call voice set and the second voiceprint feature is calculated respectively. If the similarity between the voiceprint feature of call voice 1, the voiceprint feature of call voice 2 and the second voiceprint feature is greater than or equal to the fourth preset similarity threshold, call voice 1 and call voice 2 are determined as the fourth call voice. The voiceprint feature of the fourth call voice determined in this way has a high reliability and can accurately reflect the real voiceprint feature of the speaker. In addition, the voiceprint features of the fourth call voice have the characteristics of homogeneity, such as they are all voiceprint features when the speaker speaks calmly.

[0093] Then, the similarity between the voiceprint feature of each fourth call voice and the voiceprint feature of each call voice in the candidate call voice set is calculated respectively. If the similarity between the first voiceprint feature and the voiceprint feature of call voice 1 is greater than or equal to the fifth preset similarity threshold, and the similarity between the voiceprint feature of call voice 3 and the voiceprint feature of call voice 1 is greater than or equal to the fifth preset similarity threshold, the first call voice and call voice 3 are determined as the fifth call voice. The fifth call voice thus determined not only comes from the same information source as the fourth call voice, but also covers the voiceprint features of the speaker in different states, such as voiceprint features in states of slightly angry, energetic, and relatively tired.

[0094] Furthermore, the first voiceprint feature and the voiceprint features of call voices 1 to 3 are averaged to obtain the first candidate voiceprint feature of the speaker. In this way, the first candidate voiceprint feature covers the voiceprint characteristics of the speaker in various states and is more robust. The third voiceprint feature obtained based on this update is also more robust, which is conducive to improving the accuracy of identity authentication.

[0095] It is worth noting that the above average can be either an equal-weighted average or a weighted average based on speech duration. The specific selection can be made according to actual needs, and the embodiments of the present application do not limit this.

[0096] In step A2, the first candidate voiceprint feature and the second voiceprint feature can be fused by weighted summation, that is, E t =αe t +(1-α)E t-1 . Among them, represents the third voiceprint feature, represents the first candidate voiceprint feature, represents the second voiceprint feature, and α represents the weight, which is a value between [0,1]. When α=0, the second voiceprint feature is not updated. The larger the α is, the faster the second voiceprint feature is forgotten. In practical applications, in order to ensure the effectiveness of the second voiceprint feature and avoid the second voiceprint feature being updated in special circumstances due to excessive weight, thereby affecting the accuracy of identity authentication, such as the speaker's recent cold causing the second voiceprint feature to be updated, but the voiceprint feature comparison cannot pass after the body recovers, the weight can be set to a smaller value.

[0097] When the number of second voiceprint features is at least two, all the second voiceprint features and the first candidate voiceprint features may be fused, or only some of the second voiceprint features and the first candidate voiceprint features may be fused, or only the second voiceprint features and the first candidate voiceprint features that meet the update conditions may be fused, and so on. This is not limited to the embodiments of the present application.

[0098] In order to ensure the effectiveness of the third voiceprint feature, in a more preferred manner, if the number of second voiceprint features is multiple, the above-mentioned step A2 includes: selecting a fourth voiceprint feature from multiple second voiceprint features, the fourth voiceprint feature being the second voiceprint feature with the greatest similarity to the first candidate voiceprint feature; fusing the first candidate voiceprint feature and the fourth voiceprint feature to obtain the second candidate voiceprint feature; averaging the second candidate voiceprint feature and multiple second voiceprint features to obtain the third voiceprint feature.

[0099] For example, five second voiceprint features are found from the voiceprint database, including voiceprint features 1 to 5, where voiceprint feature 1 has the greatest similarity to the first candidate voiceprint feature, then the first candidate voiceprint feature and voiceprint feature 1 are fused, for example, by weighted summation, to obtain the second candidate voiceprint feature. Then, the second candidate voiceprint feature and voiceprint features 2 to 5 are used as the third voiceprint feature.

[0100] It can be understood that in the above implementation, based on the first similarity and the information source of the second voiceprint feature, it is determined whether to update the second voiceprint feature, and in the case of determining to update the second voiceprint feature, the second voiceprint feature is updated based on the first voiceprint feature to obtain the third voiceprint feature, which can achieve real-time evaluation of the second voiceprint feature, and automatically determine whether to update the second voiceprint feature and at what frequency to update the second voiceprint feature, effectively eliminating the impact of voiceprint changes on the accuracy of identity authentication, while reducing the resource consumption caused by the second voiceprint feature of the speaker. In addition, when updating the second voiceprint feature, the first voiceprint feature is integrated to ensure that the updated third voiceprint feature and the first voice used for identity authentication are from the same information source, so as to overcome the problem of low accuracy caused by cross-information source comparison.

[0101] The embodiment of the present application shows some implementation methods of the above S206. Of course, it should be understood that the above S206 can also be implemented in other ways, and the embodiment of the present application does not limit this.

[0102] The voiceprint processing method provided by one or more embodiments of the present application extracts the first voiceprint feature of the speaker's first call speech, and queries the speaker's second voiceprint feature from the voiceprint library; the similarity between the speaker's second voiceprint feature and the first voiceprint feature can reflect whether the first voiceprint feature and the second voiceprint feature are the same person. On this basis, based on the information source of the first voiceprint feature and the information source of the second voiceprint feature, the first voiceprint feature and the second voiceprint feature are fused to obtain the speaker's third voiceprint feature, which not only makes the information source of the third voiceprint feature include the information source of the first voiceprint feature, overcoming the cross-information source comparison problem in the identity authentication process, but also effectively eliminates the impact of voiceprint changes on the accuracy of identity authentication, thereby improving the accuracy of identity authentication. In addition, the above method can be used to realize real-time and automatic update of the second voiceprint feature during a call, without the need for the speaker to register repeatedly, reducing resource consumption.

[0103] The voiceprint processing method provided in the embodiment of the present application can be applied to various call scenarios that require identity authentication. The voiceprint processing method provided in the embodiment of the present application is described below using the call center agent outbound call scenario as an example. It should be understood that the application of the voiceprint processing method provided in the embodiment of the present application to the agent outbound call scenario is only an exemplary description and should not be understood as a limitation on the application scenario of the technical solution.

[0104] like Figure 3As shown, the agent logs in to the call center system to initiate a call to the customer through the call center system. When the agent logs in to the call center system, he / she enters his / her work number. Based on the work number, the call center system queries the second voiceprint feature of the agent in the voiceprint library. If the second voiceprint feature is found, it is determined that the agent has completed the voiceprint registration, and then a prompt message is displayed to the agent to perform voiceprint authentication on the agent. If the authentication is successful, it is determined that the login is successful, and the agent is allowed to use the outbound call function of the call center system; if the authentication fails, the agent login is denied.

[0105] The specific voiceprint authentication process is as follows: the call center system displays the specified text to the agent and receives the agent's voice reading of the specified text; then, the call center system compares the voice content of the voice reading with the specified text, and extracts the voiceprint features from the voice reading and compares them with the agent's second voiceprint features. If both are consistent, the authentication is determined to be successful; otherwise, the authentication is determined to be failed. In actual applications, the agent can be repeatedly authenticated with voiceprints to improve the reliability of authentication.

[0106] If the call center system does not find the second voiceprint feature of the agent, it will prompt the agent to enter the account information and authenticate the agent based on the account information entered by the agent; if the authentication is successful, the registration prompt information will be displayed to the agent to prompt the agent to upload the voice for registering the second voiceprint feature. The call center system extracts the voiceprint feature from the voice uploaded by the agent and adds it to the voiceprint library as the second voiceprint feature of the agent (in this case, the original registered voiceprint feature). Then, enter the above voiceprint authentication process.

[0107] like Figure 4 As shown, after the agent successfully logs into the call center system, he uses the outbound call function of the call center system to initiate a call to the customer. During the call with the customer, the call center system collects the agent's call voice in real time, and queries the voiceprint library to see whether the agent has registered a second voiceprint feature whose information source is the call voice, that is, whether the information source of the agent's second voiceprint feature is the same as the information source of the first voiceprint feature of the call voice; if the two are the same, the first voiceprint feature of the call voice is extracted and compared with the second voiceprint feature of the agent for similarity. If the comparison is successful, it is determined that the agent has passed the identity authentication, thereby avoiding the situation where the agent uses his own voice to register the voiceprint feature, but is replaced by others when making an outbound call; if the comparison fails, the identity exception handling process is entered, such as using face recognition, superior authorization, etc. to authenticate the agent again. If the identity authentication is passed again, it enters Figure 5 The second voiceprint update process is shown.

[0108] If the agent has not registered the second voiceprint feature of the call voice as the information source, that is, the information source of the agent's second voiceprint feature is different from the information source of the first voiceprint feature of the call voice, then enter Figure 5The second voiceprint update process shown is to register the second voiceprint feature of the call as the information source in the voiceprint database. Then, the first voiceprint feature of the call speech is similarly compared with the second voiceprint feature to authenticate the seat.

[0109] like Figure 5 As shown, the second voiceprint feature currently available to the agent is queried in the voiceprint library, and the first similarity between the second voiceprint feature and the first voiceprint feature of the call speech is determined. Then, based on the first similarity, as well as the information source of the first voiceprint feature and the information source of the second voiceprint feature, the first voiceprint feature and the second voiceprint feature are fused to obtain a third voiceprint feature.

[0110] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0111] Based on the same inventive concept, the present application embodiment also provides a structural diagram of a voiceprint processing device. Figure 6 , is a structural diagram of a voiceprint processing device 600 provided in an embodiment of the present application. In a software implementation, the voiceprint processing device 600 may include: an acquisition module 610, a query module 620 and an update module 630.

[0112] The acquisition module 610 is used to extract a first voiceprint feature of a first call speech of a speaker.

[0113] The query module 620 is used to query the second voiceprint feature of the speaker in the voiceprint database.

[0114] The updating module 630 is used to fuse the first voiceprint feature and the second voiceprint feature based on the first similarity between the first voiceprint feature and the second voiceprint feature, the information source of the first voiceprint feature, and the information source of the second voiceprint feature to obtain a third voiceprint feature of the speaker.

[0115] In another embodiment, the acquisition module is used to:

[0116] Acquire multiple voice clips from the current call voice of the speaker;

[0117] Extract the voiceprint features of each speech segment;

[0118] The voiceprint features of the plurality of voice segments are averaged to obtain a first voiceprint feature of the first call voice.

[0119] In another embodiment, the update module is used to:

[0120] determining whether to update the second voiceprint feature based on the first similarity, the information source of the first voiceprint feature, and the information source of the second voiceprint feature;

[0121] If the second voiceprint feature is updated, the first voiceprint feature and the second voiceprint feature are fused to obtain the third voiceprint feature of the speaker.

[0122] In another embodiment, when the updating module determines whether to update the second voiceprint feature based on the first similarity, the information source of the first voiceprint feature, and the information source of the second voiceprint feature, the updating module performs the following steps:

[0123] If the information source of the first voiceprint feature is different from the information source of the second voiceprint feature, and the first similarity is greater than or equal to a first preset similarity threshold, it is determined to update the second voiceprint feature.

[0124] In another embodiment, when the updating module determines whether to update the second voiceprint feature based on the first similarity, the information source of the first voiceprint feature, and the information source of the second voiceprint feature, the updating module performs the following steps:

[0125] If the information source of the first voiceprint feature is the same as the information source of the second voiceprint feature, and the first similarity is less than a first preset similarity threshold, obtaining a first identity feature of the speaker;

[0126] authenticating the speaker based on the first identity feature;

[0127] If the speaker passes the identity authentication, it is determined to update the second voiceprint feature.

[0128] In another embodiment, when the updating module determines whether to update the second voiceprint feature based on the first similarity, the information source of the first voiceprint feature, and the information source of the second voiceprint feature, the updating module performs the following steps:

[0129] If the information source of the first voiceprint feature is the same as the information source of the second voiceprint feature, and the first similarity is greater than or equal to a first preset similarity threshold, determining the duration between the generation time point of the second voiceprint feature and the determination time point of the first similarity;

[0130] If the duration is less than a first preset duration and the first similarity is less than a second preset similarity threshold, determining to update the second voiceprint feature;

[0131] If the duration is greater than the second preset duration, or the duration is greater than or equal to the first preset duration and the first similarity is less than a second preset similarity threshold, it is determined to update the second voiceprint feature, wherein the second preset duration is greater than the first preset duration.

[0132] In another embodiment, when the updating module updates the second voiceprint feature based on the first voiceprint feature to obtain the third voiceprint feature of the speaker, the updating module performs the following steps:

[0133] Determining a first candidate voiceprint feature of the speaker based on the first voiceprint feature;

[0134] The first candidate voiceprint feature and the second voiceprint feature are fused to obtain a third voiceprint feature of the speaker.

[0135] In another embodiment, when the updating module determines the first candidate voiceprint feature of the speaker based on the first voiceprint feature, the updating module performs the following steps:

[0136] Acquire the voiceprint feature of the second call speech of the speaker;

[0137] The voiceprint feature of the second call speech and the first voiceprint feature are averaged to obtain a first candidate voiceprint feature of the speaker.

[0138] In another embodiment, when the updating module averages the voiceprint feature of the second call speech and the first voiceprint feature to obtain the first candidate voiceprint feature of the speaker, the updating module performs the following steps:

[0139] Selecting a third call voice from the second call voice, wherein a second similarity between the voiceprint feature of the third call voice and the second voiceprint feature is greater than or equal to a third preset similarity threshold;

[0140] The voiceprint feature of the third call speech and the first voiceprint feature are averaged to obtain a first candidate voiceprint feature of the speaker.

[0141] In another embodiment, when the updating module averages the voiceprint feature of the second call speech and the first voiceprint feature to obtain the first candidate voiceprint feature of the speaker, the updating module performs the following steps:

[0142] generating a candidate call voice set based on the second call voice and the first call voice;

[0143] Selecting a fourth call voice from the candidate call voice set, wherein a third similarity between the voiceprint feature of the fourth call voice and the second voiceprint feature is greater than or equal to a fourth preset similarity threshold;

[0144] Based on the voiceprint feature of the fourth call voice, selecting a fifth call voice from the candidate call voice set, wherein a fourth similarity between the voiceprint feature of the fifth call voice and the second voiceprint feature is less than the fourth preset similarity threshold, and a fifth similarity between the voiceprint feature of the fifth call voice and the fourth call voice is greater than or equal to a fifth preset similarity threshold;

[0145] The voiceprint feature of the fourth call speech and the voiceprint feature of the fifth call speech are averaged to obtain a first candidate voiceprint feature of the speaker.

[0146] In another embodiment, the number of the second voiceprint features is multiple;

[0147] When the updating module fuses the first candidate voiceprint feature and the second voiceprint feature to obtain the third voiceprint feature of the speaker, the updating module performs the following steps:

[0148] Selecting a fourth voiceprint feature from the plurality of second voiceprint features, wherein the fourth voiceprint feature is the second voiceprint feature having the greatest similarity to the first candidate voiceprint feature;

[0149] fusing the first candidate voiceprint feature and the fourth voiceprint feature to obtain a second candidate voiceprint feature;

[0150] The second candidate voiceprint feature and a plurality of the second voiceprint features are averaged to obtain a third voiceprint feature of the speaker.

[0151] Obviously, the voiceprint processing device 600 provided in the embodiment of the present application can be used as the above Figure 2 The execution body of the voiceprint processing method shown in the figure can realize the voiceprint processing device in Figure 2 The functions realized are as follows. Since the principles are the same, they will not be described in detail.

[0152] Figure 7 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present application. Figure 7 At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. The memory may include a memory, such as a high-speed random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage. Of course, the electronic device may also include hardware required for other services.

[0153] The processor, network interface and memory can be interconnected through an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0154] The memory is used to store the program. Specifically, the program may include a program code, and the program code includes a computer operation instruction. The memory may include a memory and a non-volatile memory, and provides instructions and data to the processor.

[0155] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a voiceprint processing device at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:

[0156] Extracting a first voiceprint feature of a first call speech of a speaker;

[0157] Querying a voiceprint database for a second voiceprint feature of the speaker;

[0158] Based on the first similarity between the first voiceprint feature and the second voiceprint feature, the information source of the first voiceprint feature and the information source of the second voiceprint feature, the first voiceprint feature and the second voiceprint feature are fused to obtain a third voiceprint feature of the speaker.

[0159] The above application Figure 1The method performed by the voiceprint processing device disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor or instructions in software form. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware decoding processor, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0160] The electronic device may also perform Figure 1 The method and the voiceprint processing device are implemented in Figures 1 to 5 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.

[0161] Of course, in addition to software implementation methods, the electronic device of the present application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0162] The embodiment of the present application also provides a computer-readable storage medium, which stores one or more programs, wherein the one or more programs include instructions, which, when executed by a portable electronic device including a plurality of application programs, enable the portable electronic device to execute Figure 1 The method of the embodiment shown is specifically used to perform the following operations:

[0163] Extracting a first voiceprint feature of a first call speech of a speaker;

[0164] Querying a voiceprint database for a second voiceprint feature of the speaker;

[0165] Based on the first similarity between the first voiceprint feature and the second voiceprint feature, the information source of the first voiceprint feature and the information source of the second voiceprint feature, the first voiceprint feature and the second voiceprint feature are fused to obtain a third voiceprint feature of the speaker.

[0166] An embodiment of the present application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute part or all of the steps in the voiceprint processing method provided in the embodiment of the present application.

[0167] In short, the above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0168] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0169] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0170] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0171] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

Claims

1. A voiceprint processing method, characterized in that: include: Extracting a first voiceprint feature of a first call speech of a speaker; Querying a voiceprint database for a second voiceprint feature of the speaker; Based on the first similarity between the second voiceprint feature and the first voiceprint feature, the information source of the first voiceprint feature, and the information source of the second voiceprint feature, the first voiceprint feature and the second voiceprint feature are fused to obtain a third voiceprint feature of the speaker.

2. The method according to claim 1, characterized in that The extracting the first voiceprint feature of the first call speech of the speaker includes: Acquire multiple voice segments from the first call voice of the speaker; Extract the voiceprint features of each speech segment; The voiceprint features of the plurality of voice segments are averaged to obtain a first voiceprint feature of the first call voice.

3. The method according to claim 1, characterized in that The fusing the first voiceprint feature and the second voiceprint feature based on the first similarity between the second voiceprint feature and the first voiceprint feature, and obtaining the third voiceprint feature of the speaker includes: determining whether to update the second voiceprint feature based on the first similarity, the information source of the first voiceprint feature, and the information source of the second voiceprint feature; If the second voiceprint feature is updated, the first voiceprint feature and the second voiceprint feature are fused to obtain the third voiceprint feature of the speaker.

4. The method according to claim 3, characterized in that The determining whether to update the second voiceprint feature based on the first similarity, the information source of the first voiceprint feature, and the information source of the second voiceprint feature includes: If the information source of the first voiceprint feature is the same as the information source of the second voiceprint feature, and the first similarity is greater than or equal to a first preset similarity threshold, it is determined to update the second voiceprint feature.

5. The method according to claim 3, characterized in that: The determining whether to update the second voiceprint feature based on the first similarity, the information source of the first voiceprint feature, and the information source of the second voiceprint feature includes: If the information source of the first voiceprint feature is the same as the information source of the second voiceprint feature, and the first similarity is less than a first preset similarity threshold, obtaining a first identity feature of the speaker; authenticating the speaker based on the first identity feature; If the speaker passes the identity authentication, it is determined to update the second voiceprint feature.

6. The method according to claim 3, characterized in that The determining whether to update the second voiceprint feature based on the first similarity, the information source of the first voiceprint feature, and the information source of the second voiceprint feature includes: If the information source of the first voiceprint feature is the same as the information source of the second voiceprint feature, and the first similarity is greater than or equal to a first preset similarity threshold, determining the duration between the generation time point of the second voiceprint feature and the determination time point of the first similarity; If the duration is less than a first preset duration and the first similarity is less than a second preset similarity threshold, determining to update the second voiceprint feature; If the duration is greater than the second preset duration, or the duration is greater than or equal to the first preset duration and the first similarity is less than a second preset similarity threshold, it is determined to update the second voiceprint feature, wherein the second preset duration is greater than the first preset duration.

7. The method according to claim 3, characterized in that The fusing the first voiceprint feature and the second voiceprint feature to obtain a third voiceprint feature of the speaker includes: Determining a first candidate voiceprint feature of the speaker based on the first voiceprint feature; The first candidate voiceprint feature and the second voiceprint feature are fused to obtain a third voiceprint feature of the speaker.

8. The method according to claim 7, characterized in that The determining, based on the first voiceprint feature, a first candidate voiceprint feature of the speaker includes: Acquire the voiceprint feature of the second call speech of the speaker; The voiceprint feature of the second call speech and the first voiceprint feature are averaged to obtain a first candidate voiceprint feature of the speaker.

9. The method according to claim 8, characterized in that The averaging the voiceprint feature of the second call speech and the first voiceprint feature to obtain a first candidate voiceprint feature of the speaker includes: Selecting a third call voice from the second call voice, wherein a second similarity between the voiceprint feature of the third call voice and the second voiceprint feature is greater than or equal to a third preset similarity threshold; The voiceprint feature of the third call speech and the first voiceprint feature are averaged to obtain a first candidate voiceprint feature of the speaker.

10. The method according to claim 8, characterized in that The averaging the voiceprint feature of the second call speech and the first voiceprint feature to obtain a first candidate voiceprint feature of the speaker includes: generating a candidate call voice set based on the second call voice and the first call voice; Selecting a fourth call voice from the candidate call voice set, wherein a third similarity between the voiceprint feature of the fourth call voice and the second voiceprint feature is greater than or equal to a fourth preset similarity threshold; Based on the voiceprint feature of the fourth call voice, selecting a fifth call voice from the candidate call voice set, wherein a fourth similarity between the voiceprint feature of the fifth call voice and the second voiceprint feature is less than the fourth preset similarity threshold, and a fifth similarity between the voiceprint feature of the fifth call voice and the voiceprint feature of the fourth call voice is greater than or equal to a fifth preset similarity threshold; The voiceprint feature of the fourth call speech and the voiceprint feature of the fifth call speech are averaged to obtain a first candidate voiceprint feature of the speaker.

11. The method according to claim 7, characterized in that The number of the second voiceprint features is multiple; The fusing the first candidate voiceprint feature and the second voiceprint feature to obtain the third voiceprint feature of the speaker includes: Selecting a fourth voiceprint feature from the plurality of second voiceprint features, wherein the fourth voiceprint feature is the second voiceprint feature having the greatest similarity to the first candidate voiceprint feature; fusing the first candidate voiceprint feature and the fourth voiceprint feature to obtain a second candidate voiceprint feature; The second candidate voiceprint feature and a plurality of the second voiceprint features are averaged to obtain a third voiceprint feature of the speaker.

12. A voiceprint processing device, characterized in that: include: An acquisition module, used for extracting a first voiceprint feature of a first call speech of a speaker; A query module, used for querying the second voiceprint feature of the speaker in the voiceprint library; An updating module is used to fuse the first voiceprint feature and the second voiceprint feature based on the first similarity between the second voiceprint feature and the first voiceprint feature, the information source of the first voiceprint feature, and the information source of the second voiceprint feature to obtain a third voiceprint feature of the speaker.

13. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the voiceprint processing method according to any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that: When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the voiceprint processing method as claimed in any one of claims 1 to 11.

15. A computer program product, characterized in that The computer program product comprises a non-transitory computer-readable storage medium storing a computer program, wherein the computer program is operable to cause a computer to execute part or all of the steps in the voiceprint processing method according to any one of claims 1 to 11.

Images