Voice data processing method, device, electronic device, storage medium and product

Abstract

The present application provides a method, device, electronic device, storage medium and product for processing voice data, belonging to the field of voiceprint recognition technology. The method includes: determining voice data, the voice data is the voice data to be processed; determining the crowd characteristics corresponding to the voice data, the crowd characteristics are used to represent the crowd to which the object of the voice data belongs; determining a target voice data set corresponding to the crowd characteristics from multiple voice data sets, the multiple voice data sets are obtained based on the division of crowd characteristics, and each voice data set corresponds to different crowd characteristics; determining the target voice data matching the voice data from the target voice data set, the voiceprint characteristics of the target voice data matching the voiceprint characteristics of the voice data. This method improves the processing efficiency of voice data.

Description

Technical Field

[0001] The present application relates to the field of voiceprint recognition technology, and in particular to a method, device, electronic device, storage medium and product for processing voice data. Background Art

[0002] The voice database is used to store voice data of multiple objects with known identity information, and the voice database is used to provide data support for identity recognition, etc. After collecting new voice data, the voiceprint features of the voice data are compared one by one with the voiceprint features of each voice data in the voice database to find the target voice data that matches the voice data, so as to determine the identity information corresponding to the voice data based on the identity information of the target voice data.

[0003] Since the voiceprint features of the speech data need to be compared one by one with the voiceprint features of each speech data in the speech database, and the process of voiceprint feature comparison is complicated, the comparison takes a long time, resulting in low speech data processing efficiency. Summary of the invention

[0004] The embodiments of the present application provide a method, device, electronic device, storage medium and product for processing voice data, which can improve the processing efficiency of voice data. The technical solution is as follows:

[0005] On the one hand, a method for processing speech data is provided, the method comprising:

[0006] Determining voice data, where the voice data is voice data to be processed;

[0007] Determine a crowd feature corresponding to the voice data, wherein the crowd feature is used to indicate the crowd to which the object of the voice data belongs;

[0008] Determining a target voice data set corresponding to the crowd characteristics from a plurality of voice data sets, wherein the plurality of voice data sets are obtained based on the crowd characteristics, and each voice data set corresponds to a different crowd characteristic;

[0009] Target voice data matching the voice data is determined from the target voice data set, wherein the voiceprint features of the target voice data match the voiceprint features of the voice data.

[0010] In a possible implementation, the process of determining the crowd characteristics corresponding to the voice data includes:

[0011] The voice data is input into a voiceprint recognition model, and the crowd characteristics corresponding to the voice data are output. The voiceprint recognition model is used to identify the crowd characteristics corresponding to the voice data.

[0012] In a possible implementation, there are multiple crowd features, and the voiceprint recognition model includes multiple classification task modules, each classification task module is used to obtain a crowd feature;

[0013] The step of inputting the voice data into a voiceprint recognition model and outputting crowd characteristics corresponding to the voice data includes:

[0014] The voice data is input into the voiceprint recognition model, and based on the multiple classification task modules, multiple population features corresponding to the voice data are determined, and the multiple population features are output.

[0015] In a possible implementation, the multiple population characteristics include gender and age, the multiple classification task modules include a gender classification module and an age classification module, and the step of inputting the voice data into the voiceprint recognition model and determining the multiple population characteristics corresponding to the voice data based on the multiple classification task modules includes:

[0016] The voice data is input into the voiceprint recognition model, and the gender and age corresponding to the voice data are determined based on the gender classification module and the age classification module respectively.

[0017] In a possible implementation, the training process of the voiceprint recognition model includes:

[0018] Input the sample speech data into the initial voiceprint recognition model, and output the prediction result of each classification task module on the sample speech data;

[0019] For each classification task module, comparing the prediction result with the labeling result of the sample speech data to obtain a comparison result;

[0020] Determine a first loss value corresponding to the comparison result;

[0021] Based on the first loss values ​​of the multiple classification task modules, the model parameters of the initial voiceprint recognition model are adjusted to obtain the voiceprint recognition model.

[0022] In a possible implementation, adjusting the model parameters of the initial voiceprint recognition model based on the first loss values ​​of the multiple task modules to obtain the voiceprint recognition model includes:

[0023] Based on the first loss values ​​of the multiple classification task modules, adjusting the model parameters of each classification task module respectively;

[0024] Determine the corrected second loss value of each classification task module and the weight of each classification task module at the current moment;

[0025] Taking a weighted sum of the second loss values ​​and weights of the multiple classification task modules to obtain a comprehensive loss value of the voiceprint recognition model;

[0026] Based on the comprehensive loss value, the model parameters of each classification task module are adjusted again to obtain the voiceprint recognition model.

[0027] In a possible implementation, the process of determining the weight of each classification task module at the current moment includes:

[0028] For each classification task module, determining a weight adjustment parameter of the classification task module and a measurement index value of the classification task module at the current moment;

[0029] Based on the measurement index value, determining a target logarithmic value, wherein the target logarithmic value is logarithmically related to the measurement index value;

[0030] Based on the measurement index value, determining a target difference value, wherein the target difference value is negatively correlated with the measurement index value;

[0031] Determine an exponential value with the target difference as a base and the weight adjustment parameter as an exponent;

[0032] The negative of the product of the index value and the target logarithm value is determined to obtain the weight of the classification task module.

[0033] In a possible implementation, the voiceprint recognition model further includes a voiceprint classification module, and the voiceprint classification module is used to determine target voice data that matches the voice data, and the determining the target voice data that matches the voice data from the target voice data set includes:

[0034] The voice data is input into the voiceprint recognition model, and based on the voiceprint classification module, the voiceprint features of the voice data are determined, and based on the voiceprint features, the target voice data is determined from the target voice data set.

[0035] In a possible implementation, there are multiple crowd features, and determining a target speech data set corresponding to the crowd features from multiple speech data sets includes:

[0036] Determine a speech data set corresponding to each crowd feature respectively, wherein the speech data set includes at least one speech data corresponding to the crowd feature;

[0037] The intersection of the speech data sets corresponding to the plurality of population characteristics is obtained to obtain the target speech data set.

[0038] In another aspect, a device for processing speech data is provided, the device comprising:

[0039] A first determining module, used to determine voice data, where the voice data is voice data to be processed;

[0040] A second determination module is used to determine a crowd feature corresponding to the voice data, where the crowd feature is used to indicate the crowd to which the object of the voice data belongs;

[0041] A third determination module is used to determine a target voice data set corresponding to the crowd feature from a plurality of voice data sets, wherein the plurality of voice data sets are obtained based on the crowd features, and each voice data set corresponds to a different crowd feature;

[0042] The fourth determination module is used to determine target voice data matching the voice data from the target voice data set, wherein the voiceprint features of the target voice data match the voiceprint features of the voice data.

[0043] In a possible implementation, the second determining module includes:

[0044] The first input-output unit is used to input the voice data into a voiceprint recognition model and output crowd characteristics corresponding to the voice data. The voiceprint recognition model is used to identify the crowd characteristics corresponding to the voice data.

[0045] In a possible implementation, there are multiple crowd features, and the voiceprint recognition model includes multiple classification task modules, each classification task module is used to obtain a crowd feature;

[0046] The first input-output unit comprises:

[0047] The input-output subunit is used to input the voice data into the voiceprint recognition model, determine multiple population features corresponding to the voice data based on the multiple classification task modules, and output the multiple population features.

[0048] In one possible implementation, the multiple population characteristics include gender and age, the multiple classification task modules include a gender classification module and an age classification module, and the input-output subunit is used to input the voice data into the voiceprint recognition model, and determine the gender and age corresponding to the voice data based on the gender classification module and the age classification module, respectively.

[0049] In a possible implementation manner, the device further includes:

[0050] An input-output module, used to input the sample voice data into the initial voiceprint recognition model, and output the prediction result of each classification task module on the sample voice data;

[0051] A comparison module, used for comparing the prediction result with the labeling result of the sample speech data for each classification task module to obtain a comparison result;

[0052] A fifth determining module, used to determine a first loss value corresponding to the comparison result;

[0053] An adjustment module is used to adjust the model parameters of the initial voiceprint recognition model based on the first loss values ​​of the multiple classification task modules to obtain the voiceprint recognition model.

[0054] In a possible implementation, the adjustment module includes:

[0055] A first adjustment unit, configured to adjust the model parameters of each classification task module based on the first loss values ​​of the multiple classification task modules;

[0056] A first determining unit, used to determine the corrected second loss value of each classification task module and the weight of each classification task module at a current moment;

[0057] A second determining unit, configured to perform a weighted summation of the second loss values ​​and weights of the plurality of classification task modules to obtain a comprehensive loss value of the voiceprint recognition model;

[0058] The second adjustment unit is used to adjust the model parameters of each classification task module again based on the comprehensive loss value to obtain the voiceprint recognition model.

[0059] In a possible implementation manner, the first determining unit is configured to:

[0060] For each classification task module, determining a weight adjustment parameter of the classification task module and a measurement index value of the classification task module at the current moment;

[0061] Based on the measurement index value, determining a target logarithmic value, wherein the target logarithmic value is logarithmically related to the measurement index value;

[0062] Based on the measurement index value, determining a target difference value, wherein the target difference value is negatively correlated with the measurement index value;

[0063] Determine an exponential value with the target difference as a base and the weight adjustment parameter as an exponent;

[0064] The negative of the product of the index value and the target logarithm value is determined to obtain the weight of the classification task module.

[0065] In one possible implementation, the voiceprint recognition model also includes a voiceprint classification module, and the voiceprint classification module is used to determine target voice data that matches the voice data. The fourth determination module is used to input the voice data into the voiceprint recognition model, determine the voiceprint features of the voice data based on the voiceprint classification module, and determine the target voice data from the target voice data set based on the voiceprint features.

[0066] In a possible implementation, there are multiple population characteristics, and the second determining module is used to:

[0067] Determine a speech data set corresponding to each crowd feature respectively, wherein the speech data set includes at least one speech data corresponding to the crowd feature;

[0068] The intersection of the speech data sets corresponding to the plurality of population characteristics is obtained to obtain the target speech data set.

[0069] On the other hand, an electronic device is provided, comprising one or more processors and one or more memories, wherein at least one instruction is stored in the one or more memories, and the at least one instruction is loaded and executed by the one or more processors to implement the method for processing voice data described in any of the above-mentioned implementation modes.

[0070] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored in the computer-readable storage medium, and the at least one instruction is loaded and executed by a processor to implement the method for processing voice data described in any of the above implementations.

[0071] On the other hand, a computer program product is provided, the computer program product comprising at least one program code, the at least one program code being loaded and executed by a processor to implement the method for processing voice data described in any of the above implementations.

[0072] The beneficial effects of the technical solution provided by the embodiments of the present application include at least:

[0073] An embodiment of the present application provides a method for processing voice data. Since the method can determine a target voice data set corresponding to the population characteristics in multiple voice data sets based on the population characteristics corresponding to the voice data to be processed, the comparison range of the voice data is narrowed; and then the target voice data matching the voice data is determined from the target voice data set, which reduces the difficulty of comparison, reduces the time spent on comparison, and thus improves the processing efficiency of the voice data. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0075] Figure 1 It is a schematic diagram of an implementation environment provided by an embodiment of the present application;

[0076] Figure 2 is a flow chart of a method for processing voice data provided by an embodiment of the present application;

[0077] Figure 3 is a flow chart of a method for processing voice data provided by an embodiment of the present application;

[0078] Figure 4 It is a flow chart of a method for training a voiceprint recognition model provided in an embodiment of the present application;

[0079] Figure 5 is a schematic diagram of a multi-task learning deep network provided in an embodiment of the present application;

[0080] Figure 6 is a block diagram of a voice data processing device provided in an embodiment of the present application;

[0081] Figure 7 is a block diagram of a terminal provided in an embodiment of the present application;

[0082] Figure 8 This is a block diagram of a server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0083] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0084] The terms "first", "second", "third" and "fourth" etc. in the specification and claims of the present application and the drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0085] The present application embodiment provides an implementation environment of a method for processing voice data, see Figure 1, the implementation environment includes an electronic device. In some embodiments, the electronic device includes a terminal 10, on which a voiceprint recognition model is stored, and the voiceprint recognition model is used to determine target voice data that matches the voice data, and the target voice data is voice data with known identity information. After the terminal 10 obtains the voice data to be processed, based on the voiceprint recognition model, it determines the target voice data that matches the voice data, and determines the identity information corresponding to the voice data based on the identity information of the target voice data. The terminal 10 is at least one of a mobile phone, a tablet computer, a PC (Personal Computer) device, and the like.

[0086] In some embodiments, the electronic device includes a server 20, and a voiceprint recognition model is stored on the server 20. The voiceprint recognition model is used to determine target voice data that matches the voice data, and the target voice data is voice data with known identity information. After the server 20 obtains the voice data to be processed, it determines the target voice data that matches the voice data based on the voiceprint recognition model, and determines the identity information corresponding to the voice data based on the identity information of the target voice data. The server 20 can be at least one of a server, a server cluster consisting of multiple servers, a cloud server, a cloud computing platform, and a virtualization center.

[0087] In some embodiments, the electronic device includes a terminal 10 and a server 20, and a target application providing voiceprint recognition is installed on the terminal 10, and the terminal 10 can realize functions such as data transmission and information interaction with the server 20 through the target application. A voiceprint recognition model is stored on the server 20, and the voiceprint recognition model is used to determine target voice data that matches the voice data, and the target voice data is voice data with known identity information. The terminal 10 obtains the voice data to be processed based on the target application, and sends the voice data to the server 20. After receiving the voice data, the server 20 determines the target voice data that matches the voice data based on the voiceprint recognition model, and determines the identity information corresponding to the voice data based on the identity information of the target voice data. In the embodiment of the present application, the electronic device includes a terminal 10 and a server 20 for illustration.

[0088] The present application provides a method for processing voice data. Figure 2 , methods include:

[0089] Step 201: Determine voice data, where the voice data is voice data to be processed;

[0090] Step 202: Determine the crowd characteristics corresponding to the voice data, where the crowd characteristics are used to indicate the crowd to which the object of the voice data belongs;

[0091] Step 203: determining a target voice data set corresponding to a crowd feature from a plurality of voice data sets, wherein the plurality of voice data sets are obtained based on the crowd features, and each voice data set corresponds to a different crowd feature;

[0092] Step 204: Determine target voice data that matches the voice data from the target voice data set, and the voiceprint features of the target voice data match the voiceprint features of the voice data.

[0093] In a possible implementation, the process of determining the crowd characteristics corresponding to the voice data includes:

[0094] The voice data is input into the voiceprint recognition model, and the crowd characteristics corresponding to the voice data are output. The voiceprint recognition model is used to identify the crowd characteristics corresponding to the voice data.

[0095] In a possible implementation, there are multiple crowd features, and the voiceprint recognition model includes multiple classification task modules, each classification task module is used to obtain a crowd feature;

[0096] Input the voice data into the voiceprint recognition model, and output the crowd characteristics corresponding to the voice data, including:

[0097] The voice data is input into the voiceprint recognition model, and based on multiple classification task modules, multiple population characteristics corresponding to the voice data are determined, and multiple population characteristics are output.

[0098] In a possible implementation, the multiple population characteristics include gender and age, the multiple classification task modules include a gender classification module and an age classification module, the voice data is input into the voiceprint recognition model, and based on the multiple classification task modules, multiple population characteristics corresponding to the voice data are determined, including:

[0099] The voice data is input into the voiceprint recognition model, and the gender and age corresponding to the voice data are determined based on the gender classification module and the age classification module respectively.

[0100] In a possible implementation, the training process of the voiceprint recognition model includes:

[0101] Input the sample speech data into the initial voiceprint recognition model, and output the prediction results of each classification task module on the sample speech data;

[0102] For each classification task module, compare the prediction result with the labeled result of the sample speech data to obtain a comparison result;

[0103] Determine a first loss value corresponding to the comparison result;

[0104] Based on the first loss values ​​of multiple classification task modules, the model parameters of the initial voiceprint recognition model are adjusted to obtain the voiceprint recognition model.

[0105] In a possible implementation, based on the first loss values ​​of the multiple task modules, the model parameters of the initial voiceprint recognition model are adjusted to obtain the voiceprint recognition model, including:

[0106] Based on the first loss values ​​of the multiple classification task modules, adjusting the model parameters of each classification task module respectively;

[0107] Determine the corrected second loss value of each classification task module and the weight of each classification task module at the current moment;

[0108] The second loss values ​​and weights of multiple classification task modules are weighted and summed to obtain the comprehensive loss value of the voiceprint recognition model;

[0109] Based on the comprehensive loss value, the model parameters of each classification task module are adjusted again to obtain the voiceprint recognition model.

[0110] In a possible implementation, the process of determining the weight of each classification task module at the current moment includes:

[0111] For each classification task module, determine the weight adjustment parameter of the classification task module and the measurement index value of the classification task module at the current moment;

[0112] Based on the measurement indicator value, a target logarithmic value is determined, and the target logarithmic value is in a logarithmic relationship with the measurement indicator value;

[0113] Based on the measurement index value, a target difference is determined, and the target difference is negatively correlated with the measurement index value;

[0114] Determine an index value with the target difference as the base and the weight adjustment parameter as the exponent;

[0115] The negative of the product of the index value and the target logarithm value is determined to obtain the weight of the classification task module.

[0116] In a possible implementation, the voiceprint recognition model further includes a voiceprint classification module, which is used to determine target voice data that matches the voice data, input the voice data into the voiceprint recognition model, and output the target voice data, including:

[0117] The voice data is input into the voiceprint recognition model, and based on the voiceprint classification module, the voiceprint features of the voice data are determined. Based on the voiceprint features, the target voice data is determined from the target voice data set.

[0118] In a possible implementation, there are multiple crowd features, and determining a target speech data set corresponding to the crowd features from multiple speech data sets includes:

[0119] Determine a speech data set corresponding to each crowd feature respectively, the speech data set including at least one speech data corresponding to the crowd feature;

[0120] The intersection of the speech data sets corresponding to the characteristics of multiple groups of people is obtained to obtain the target speech data set.

[0121] The present application provides a method for processing voice data. Figure 3 , methods include:

[0122] Step 301: The terminal determines voice data and sends the voice data to the server.

[0123] The voice data is the voice data to be processed; the voice data to be processed is the newly collected voice data for which identity information needs to be determined. The voice data can be the voice data to be processed acquired by the terminal in real time, or can be the voice data to be processed acquired by the terminal from the outside in advance. Optionally, the voice data is a voice segment containing only a voice signal or a voice segment containing a voice signal and interference noise.

[0124] When the terminal determines that the voice data includes a voice signal, the terminal determines that the voice data is voice data to be processed and sends the voice data to the server. In one implementation, after the terminal obtains the voice data in real time, the terminal sends the voice data to the server in real time. In another implementation, the terminal periodically sends the voice data to the server; optionally, the terminal packages the voice data obtained in the periodic time period and sends it to the server.

[0125] Step 302: The server receives voice data.

[0126] In some embodiments, when the server receives multiple voice data, the server processes each voice data in turn based on the order of the multiple voice data; the server processes the voice data with the highest order first and stores the remaining voice data.

[0127] Step 303: The server determines the crowd characteristics corresponding to the voice data.

[0128] The crowd feature is used to indicate the crowd to which the object of the voice data belongs. Optionally, the crowd feature is at least one of gender, age, accent, etc. In one implementation, the server inputs the voice data into a voiceprint recognition model and outputs the crowd feature corresponding to the voice data. The voiceprint recognition model is used to identify the crowd feature corresponding to the voice data.

[0129] In some embodiments, there are multiple crowd features, and the voiceprint recognition model includes multiple classification task modules, each of which is used to obtain a crowd feature. In this embodiment, the server inputs the voice data into the voiceprint recognition model, determines multiple crowd features corresponding to the voice data based on multiple classification task modules, and outputs multiple crowd features. For example, the multiple crowd features include gender and age, and the multiple classification task modules include a gender classification module and an age classification module. Then, the server inputs the voice data into the voiceprint recognition model, and determines the gender and age corresponding to the voice data based on the gender classification module and the age classification module, respectively.

[0130] It should be noted that since the gender, age and other crowd characteristics corresponding to the voice data are more obvious and easy to determine, the process of determining the crowd characteristics corresponding to the voice data is more efficient. In this way, determining the target voice data in the voice data set corresponding to the crowd characteristics will reduce the scope of comparison, thereby improving the efficiency of determining the target voice data.

[0131] Step 304: The server determines a target voice data set corresponding to the crowd characteristics from multiple voice data sets.

[0132] Among them, multiple voice data sets are obtained based on population characteristics, and each voice data set corresponds to different population characteristics. It should be noted that each voice data set includes at least one voice data, and each voice data in each voice data set is voice data with known identity information.

[0133] In one implementation, the number of population characteristics is 1, and multiple voice data sets are respectively obtained based on the 1 population characteristic; optionally, the population characteristic is gender, then the multiple voice data sets include two voice data sets, and the population characteristics corresponding to the two voice data sets are female and male, respectively.

[0134] In another implementation, there are multiple population features, and multiple voice data sets are obtained based on the multiple population features; optionally, the multiple population features are gender and accent, and each voice data set in the multiple voice data sets corresponds to two population features, namely, gender and accent. If the gender includes male and female, and the accent includes a southern accent and a northern accent, then the multiple voice data sets include 4 voice data sets, and the corresponding population features are female with a southern accent, female with a northern accent, male with a southern accent, and male with a northern accent. In this implementation, multiple voice data corresponding to multiple population features are grouped into the same voice data set in advance, so that after determining the multiple population features corresponding to the voice data, the target voice data set corresponding to the multiple population features can be directly found, with high accuracy.

[0135] In another implementation, there are multiple crowd features, and multiple voice data are obtained based on the multiple crowd features; the server determines the voice data set corresponding to each crowd feature, and the voice data set includes at least one voice data corresponding to the crowd feature. The server obtains the intersection of the voice data sets corresponding to the multiple crowd features to obtain the target voice data set.

[0136] Optionally, the multiple population characteristics include gender and accent, and the multiple voice data sets include four voice data sets corresponding to male, female, southern accent, and northern accent, respectively. After the server determines that the population characteristic of the voice data is female and southern accent, the voice data set corresponding to female is intersected with the voice data set corresponding to southern accent to obtain the target voice data set corresponding to the population characteristic of female and southern accent. In this way, the process of intersecting all the multiple voice data sets corresponding to all the population characteristics in advance is avoided, and only the voice data sets corresponding to each population characteristic of the voice data to be processed need to be intersected, which saves resources and improves processing efficiency.

[0137] Step 305: The server determines target voice data that matches the voice data from the target voice data set, and the voiceprint features of the target voice data match the voiceprint features of the voice data.

[0138] In an embodiment of the present application, the voiceprint recognition model is also used to determine the target voice data that matches the voice data. In one implementation, the server inputs the voice data into the voiceprint recognition model and outputs the target voice data. The voiceprint recognition model also includes a voiceprint classification module, which is used to determine the target voice data that matches the voice data. In this implementation, the server inputs the voice data into the voiceprint recognition model, determines the voiceprint features of the voice data based on the voiceprint classification module, and determines the target voice data from the target voice data set based on the voiceprint features.

[0139] The voiceprint features of the speech data to be processed and the voiceprint features of each speech data in the target speech data set are both feature vectors; the voiceprint features of the speech data to be processed are represented as the first feature vector, and the voiceprint features of each speech data in the target speech data set are represented as the second feature vector. After the server obtains the second feature vector of each speech data in the target speech data set, it calculates the similarity between the first feature vector and the plurality of second feature vectors respectively to obtain a plurality of corresponding similarities, and when the highest similarity among the plurality of similarities is greater than a preset threshold, the speech data corresponding to the highest similarity is used as the target speech data.

[0140] In one embodiment, the voiceprint classification module is a classification task module in parallel with multiple classification task modules, and is the main task classification module of the voiceprint recognition model, with a higher weight than other classification task modules, and the voiceprint classification module shares the output results of the previous multiple classification task modules. In this implementation, the voiceprint classification module determines the target voice data that matches the voice data based on the target voice data set corresponding to the crowd characteristics determined by the previous multiple classification task modules.

[0141] In one implementation, the voiceprint recognition model is also used to determine the target voice data set corresponding to the crowd characteristics, and the voiceprint recognition model stores multiple voice data sets corresponding to the crowd characteristics. In this implementation, after the server inputs the voice data into the voiceprint recognition model, the voice data is first processed by multiple classification task modules to obtain the target voice data set, and then processed by the voiceprint classification module to directly output the target voice data matching the voice data. In this way, through the voiceprint recognition model, it is possible to first determine the crowd characteristics to which the voice data belongs based on multiple auxiliary classification task modules such as the gender classification module and the age classification model, and then determine the target voice data matching the voice data from the target voice data set corresponding to the crowd characteristics, thereby reducing the comparison range of the voice data and improving the efficiency of determining the target voice data.

[0142] Step 306: The server sends target identity information to the terminal, where the target identity information is identity information corresponding to the target voice data.

[0143] In one implementation, each voice data in the target voice data set carries an identity tag for indicating the identity information corresponding to the voice data. Optionally, the target identity information includes the name, ID number, address, contact information, etc. of the object corresponding to the target identity information.

[0144] Step 307: The terminal receives the target identity information.

[0145] After receiving the target identity information, the terminal outputs the target identity information. In one implementation, the terminal is equipped with a display screen, and the target identity information can be displayed on the display screen. In another implementation, the terminal is equipped with a player, and the target identity information can be played on the player.

[0146] In the embodiment of the present application, by using a multi-task learning deep network to cluster the voice data in terms of gender and age, the retrieval time of the target voice data that actually matches the voice data is reduced, and by sharing the output results of the voiceprint classification module with multiple classification task modules such as the gender classification module and the age classification module, the mutual complementation of multiple classification task modules is achieved, and the model effect is improved; in this way, since the gender, age and other crowd characteristics are added to assist in the recognition of voiceprints, the accuracy of voiceprint recognition is improved, that is, the accuracy of voice data processing is improved. In some embodiments, by processing voice data using the method provided in the embodiment of the present application, the retrieval time for determining the target voice data that matches the voice data can be reduced by 25%, and the accuracy of voiceprint recognition can be improved by 30%, that is, the accuracy of voice data processing is improved.

[0147] An embodiment of the present application provides a method for processing voice data. Since the method can determine a target voice data set corresponding to the population characteristics in multiple voice data sets based on the population characteristics corresponding to the voice data to be processed, the comparison range of the voice data is narrowed; and then the target voice data matching the voice data is determined from the target voice data set, which reduces the difficulty of comparison, reduces the time spent on comparison, and thus improves the processing efficiency of the voice data.

[0148] The present application also provides a method for training a voiceprint recognition model. Figure 4 , the method comprising:

[0149] Step 401: The electronic device inputs sample voice data into an initial voiceprint recognition model, and outputs the prediction result of each classification task module on the sample voice data.

[0150] The sample speech data has been annotated with multiple labels, which are respectively used to mark the correct classification results of the sample speech data corresponding to each classification task module. For example, the multiple classification task modules include a gender classification module and an age classification module, and the sample speech data includes a gender label and an age label. The gender label includes at least one of male and female, and the age label includes at least one of multiple age groups such as 0-10 years old, 10-20 years old, 20-30 years old, 30-60 years old, etc.

[0151] It should be noted that the initial voiceprint recognition model is a model with preliminary training parameters, which are parameters that have not been trained by the model. Since the parameters of the initial voiceprint recognition model have not been trained, the prediction results of the sample speech data output by it are different from the labeling results of the labels. In the embodiment of the present application, the electronic device performs model training on the initial voiceprint recognition model based on a multi-task learning deep network. Figure 5 , Figure 5This is a schematic diagram of the multi-task learning deep network corresponding to the initial voiceprint recognition model, including an input layer, a shared layer and a task layer; wherein the input layer is used to input voice data, the shared layer is used to generate underlying features shared by multiple classification task modules, the task layer includes multiple parallel classification task modules, and multiple task classification modules are trained in parallel and share the classification results obtained by different classification task modules.

[0152] Step 402: The electronic device compares the prediction result with the labeling result of the sample speech data for each classification task module to obtain a comparison result.

[0153] The comparison result is used to indicate whether the prediction result matches the marked result. The comparison result includes the error value between the prediction result and the marked result. If the error value is 0, it means that the prediction result is the same as the marked result, and the comparison result indicates that the prediction result matches the marked result. If the error value is not 0, it means that the prediction result does not match the marked result, and the comparison result indicates that the prediction result does not match the marked result. For example, for sample speech data A, its marked result is 10-20 years old, and the prediction result is 20-30 years old, then the error value is not 0, the prediction result is different from the marked result, and the comparison result indicates that the prediction result does not match the marked result.

[0154] Step 403: The electronic device determines a first loss value corresponding to the comparison result.

[0155] In one implementation, the comparison result is an error value between the marked result and the predicted result.

[0156] Step 404: The electronic device adjusts the model parameters of the initial voiceprint recognition model based on the first loss values ​​of the multiple classification task modules to obtain the voiceprint recognition model.

[0157] This step includes the following steps (1)-(4):

[0158] (1) The electronic device adjusts the model parameters of each classification task module based on the first loss values ​​of the multiple classification task modules.

[0159] Among them, the first loss value of each classification task module represents the accuracy corresponding to the classification task module. The first loss value is involved in back propagation to adjust the entire multi-task learning deep network, and then the classification task module can be corrected.

[0160] (2) The electronic device determines the corrected second loss value of each classification task module and the weight of each classification task module at the current moment.

[0161] Among them, the electronic device determines the weight of each classification task module at the current moment, including: the electronic device determines the weight adjustment parameter of the classification task module and the measurement index value of the classification task module at the current moment for each classification task module; the electronic device determines the target logarithmic value based on the measurement index value, and the target logarithmic value is in a logarithmic relationship with the measurement index value; the electronic device determines the target difference value based on the measurement index value, and the target difference value is negatively correlated with the measurement index value; the electronic device determines an exponential value with the target difference value as the base and the weight adjustment parameter as the exponent; the electronic device determines the negative of the product of the exponential value and the target logarithmic value to obtain the weight of the classification task module. In one implementation, the target logarithmic value is the logarithmic value of the measurement index value with 10 as the base, and the target difference value is the difference between the value 1 and the measurement index value. The electronic device determines the weight of each classification task module at the current moment by the following formula 1.

[0162] Formula 1:

[0163] Among them, w i (t) represents the weight of each classification task module at the current moment, and t is the time. i is the weight adjustment parameter of classification task module i, used to adjust the weight of classification task module i, γ i ∈[1,2].k i (t) represents the measurement index value of each classification task module at the current moment, that is, the measurement index of the classification difficulty of classification task module i. It should be noted that k i (t) refers to KPI (Key Performance Indicator), k i (t)∈[0,1],k i (t) is related to the classification accuracy of the classification task module, and k i (t) is inversely proportional to the training difficulty of the classification task module. The higher the accuracy, the lower the training difficulty. i (t) is larger; for example, the multiple classification task modules are gender classification module, age classification module and voiceprint classification module, the training difficulty of gender classification module, age classification module and voiceprint classification module increases successively, and the accuracy decreases successively, then k i (t) increases successively.

[0164] It should be noted that the measurement index value of each classification task module changes over time. By determining the weight of the classification task module at the current moment through the measurement index value at the current moment, the weight of the classification task module can be dynamically adjusted, and by adjusting the measurement index value, the weight of the classification task module can be adjusted, so that the classification task module that is more difficult to train has a higher weight, so as to improve the classification accuracy of the classification task module that is more difficult to train, thereby improving the overall accuracy of the voiceprint recognition model. Optionally, a higher weight is set for the voiceprint classification module corresponding to the main classification task.

[0165] (3) The electronic device performs weighted summation of the second loss values ​​and weights of multiple classification task modules to obtain a comprehensive loss value of the voiceprint recognition model.

[0166] The comprehensive loss value of the voiceprint recognition model is obtained by the following formula 2.

[0167] Formula 2: L = ∑ i w i *L i

[0168] Among them, L i represents the second loss value of the i-th task, w i Represents the weight of the i-th task.

[0169] (4) Based on the comprehensive loss value, the electronic device adjusts the model parameters of each classification task module again to obtain a voiceprint recognition model.

[0170] The comprehensive loss value is used to characterize the overall accuracy of the voiceprint recognition model. When the comprehensive loss value is constantly changing, the electronic device repeats steps (1)-(3) to continuously adjust the model parameters of each classification task module until the comprehensive loss value of the voiceprint recognition model no longer decreases after each classification task module of the voiceprint recognition model has been adjusted for multiple consecutive iterations. The training is then terminated, and the training parameters with the minimum comprehensive loss value are determined as the target training parameters. The voiceprint recognition model corresponding to the target training parameters is used as the final voiceprint recognition model. The target training parameters include the model parameters of each classification task module.

[0171] In an embodiment of the present application, by performing multi-task learning training on the voiceprint recognition model, the population characteristics such as gender and age of the voice data to be processed can be determined based on the voiceprint recognition model; and then the target voice data matching the voice data is determined from the voice data set corresponding to the population characteristics, which can improve the efficiency of determining the target voice data.

[0172] The present application also provides a device for processing voice data. Figure 6 , the device comprises:

[0173] A first determining module 601 is used to determine voice data, where the voice data is the voice data to be processed;

[0174] A second determination module 602 is used to determine the crowd characteristics corresponding to the voice data, where the crowd characteristics are used to indicate the crowd to which the object of the voice data belongs;

[0175] The third determination module 603 is used to determine a target speech data set corresponding to a crowd feature from a plurality of speech data sets, wherein the plurality of speech data sets are obtained based on the crowd features, and each speech data set corresponds to a different crowd feature;

[0176] The fourth determination module 604 is used to determine the target voice data that matches the voice data from the target voice data set, and the voiceprint features of the target voice data match the voiceprint features of the voice data.

[0177] In a possible implementation, the second determining module 602 includes:

[0178] The first input-output unit is used to input the voice data into the voiceprint recognition model and output the crowd characteristics corresponding to the voice data. The voiceprint recognition model is used to identify the crowd characteristics corresponding to the voice data.

[0179] In a possible implementation, there are multiple crowd features, and the voiceprint recognition model includes multiple classification task modules, each classification task module is used to obtain a crowd feature;

[0180] The first input-output unit comprises:

[0181] The input-output subunit is used to input the voice data into the voiceprint recognition model, determine the multiple population characteristics corresponding to the voice data based on multiple classification task modules, and output the multiple population characteristics.

[0182] In one possible implementation, the multiple population characteristics include gender and age, the multiple classification task modules include a gender classification module and an age classification module, and the input and output sub-units are used to input the voice data into the voiceprint recognition model, and determine the gender and age corresponding to the voice data based on the gender classification module and the age classification module respectively.

[0183] In a possible implementation, the device further includes:

[0184] The input and output module is used to input the sample voice data into the initial voiceprint recognition model and output the prediction results of each classification task module on the sample voice data;

[0185] A comparison module, used for comparing the prediction result with the labeling result of the sample speech data for each classification task module to obtain a comparison result;

[0186] A fifth determining module, used to determine a first loss value corresponding to the comparison result;

[0187] The adjustment module is used to adjust the model parameters of the initial voiceprint recognition model based on the first loss values ​​of multiple classification task modules to obtain the voiceprint recognition model.

[0188] In a possible implementation, the adjustment module includes:

[0189] A first adjustment unit, configured to adjust the model parameters of each classification task module based on the first loss values ​​of the multiple classification task modules;

[0190] A first determining unit, used to determine the corrected second loss value of each classification task module and the weight of each classification task module at a current moment;

[0191] A second determination unit is used to perform weighted summation of the second loss values ​​and weights of the multiple classification task modules to obtain a comprehensive loss value of the voiceprint recognition model;

[0192] The second adjustment unit is used to adjust the model parameters of each classification task module again based on the comprehensive loss value to obtain a voiceprint recognition model.

[0193] In a possible implementation manner, the first determining unit is configured to:

[0194] For each classification task module, determine the weight adjustment parameter of the classification task module and the measurement index value of the classification task module at the current moment;

[0195] Based on the measurement indicator value, a target logarithmic value is determined, and the target logarithmic value is in a logarithmic relationship with the measurement indicator value;

[0196] Based on the measurement index value, a target difference is determined, and the target difference is negatively correlated with the measurement index value;

[0197] Determine an index value with the target difference as the base and the weight adjustment parameter as the exponent;

[0198] The negative of the product of the index value and the target logarithm value is determined to obtain the weight of the classification task module.

[0199] In one possible implementation, the voiceprint recognition model includes a voiceprint classification module, which is used to determine target voice data that matches the voice data. The fourth determination module 604 is used to input the voice data into the voiceprint recognition model, determine the voiceprint features of the voice data based on the voiceprint classification module, and determine the target voice data from the target voice data set based on the voiceprint features.

[0200] In a possible implementation, there are multiple crowd characteristics, and the second determination module 602 is used to:

[0201] Determine a speech data set corresponding to each crowd feature respectively, the speech data set including at least one speech data corresponding to the crowd feature;

[0202] The intersection of the speech data sets corresponding to the characteristics of multiple groups of people is obtained to obtain the target speech data set.

[0203] Figure 7 The structure block diagram of a terminal 700 provided by an exemplary embodiment of the present application is shown. The terminal 700 may be a portable mobile terminal, such as a smart phone, a tablet computer, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer or a desktop computer. The terminal 700 may also be called a user device, a portable terminal, a laptop terminal, a desktop terminal or other names.

[0204] Typically, the terminal 700 includes a processor 701 and a memory 702 .

[0205] The processor 701 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 701 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 701 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 701 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 701 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0206] The memory 702 may include one or more computer-readable storage media, which may be non-transitory. The memory 702 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 702 is used to store at least one instruction, which is used to be executed by the processor 701 to implement the method for processing voice data provided in the method embodiment of the present application.

[0207] In some embodiments, the terminal 700 may further optionally include: a peripheral device interface 703 and at least one peripheral device. The processor 701, the memory 702 and the peripheral device interface 703 may be connected via a bus or a signal line. Each peripheral device may be connected to the peripheral device interface 703 via a bus, a signal line or a circuit board. Specifically, the peripheral device includes: at least one of a radio frequency circuit 704, a display screen 705, a camera assembly 706, an audio circuit 707, a positioning assembly 708 and a power supply 709.

[0208] The peripheral device interface 703 may be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 701 and the memory 702. In some embodiments, the processor 701, the memory 702, and the peripheral device interface 703 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 701, the memory 702, and the peripheral device interface 703 may be implemented on a separate chip or circuit board, which is not limited in this embodiment.

[0209] The radio frequency circuit 704 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 704 communicates with communication networks and other communication devices through electromagnetic signals. The radio frequency circuit 704 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals into electrical signals. Optionally, the radio frequency circuit 704 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, etc. The radio frequency circuit 704 can communicate with other terminals through at least one wireless communication protocol. The wireless communication protocol includes but is not limited to: the World Wide Web, a metropolitan area network, an intranet, various generations of mobile communication networks (2G, 3G, 4G and 5G), a wireless local area network and / or a WiFi (Wireless Fidelity) network. In some embodiments, the radio frequency circuit 704 may also include circuits related to NFC (Near Field Communication), which is not limited in this application.

[0210] The display screen 705 is used to display a UI (User Interface). The UI may include graphics, text, icons, videos, and any combination thereof. When the display screen 705 is a touch display screen, the display screen 705 also has the ability to collect touch signals on the surface or above the surface of the display screen 705. The touch signal can be input to the processor 701 as a control signal for processing. At this time, the display screen 705 can also be used to provide virtual buttons and / or virtual keyboards, also known as soft buttons and / or soft keyboards. In some embodiments, the display screen 705 can be one, arranged on the front panel of the terminal 700; in other embodiments, the display screen 705 can be at least two, respectively arranged on different surfaces of the terminal 700 or in a folding design; in other embodiments, the display screen 705 can be a flexible display screen, arranged on a curved surface or a folding surface of the terminal 700. Even, the display screen 705 can also be set to a non-rectangular irregular shape, that is, a special-shaped screen. The display screen 705 can be made of materials such as LCD (Liquid Crystal Display), OLED (Organic Light-Emitting Diode, organic light-emitting diode).

[0211] The camera assembly 706 is used to capture images or videos. Optionally, the camera assembly 706 includes a front camera and a rear camera. Typically, the front camera is arranged on the front panel of the terminal, and the rear camera is arranged on the back of the terminal. In some embodiments, there are at least two rear cameras, which are any one of a main camera, a depth of field camera, a wide-angle camera, and a telephoto camera, so as to realize the fusion of the main camera and the depth of field camera to realize the background blur function, the fusion of the main camera and the wide-angle camera to realize panoramic shooting and VR (Virtual Reality) shooting function or other fusion shooting functions. In some embodiments, the camera assembly 706 may also include a flash. The flash can be a monochrome temperature flash or a dual-color temperature flash. A dual-color temperature flash refers to a combination of a warm light flash and a cold light flash, which can be used for light compensation at different color temperatures.

[0212] The audio circuit 707 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, and convert the sound waves into electrical signals and input them into the processor 701 for processing, or input them into the radio frequency circuit 704 to achieve voice communication. For the purpose of stereo acquisition or noise reduction, there may be multiple microphones, which are respectively arranged at different parts of the terminal 700. The microphone may also be an array microphone or an omnidirectional acquisition microphone. The speaker is used to convert the electrical signal from the processor 701 or the radio frequency circuit 704 into sound waves. The speaker may be a traditional film speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can not only convert the electrical signal into sound waves audible to humans, but also convert the electrical signal into sound waves inaudible to humans for purposes such as ranging. In some embodiments, the audio circuit 707 may also include a headphone jack.

[0213] Positioning component 708 is used to locate the current geographical location of terminal 700 to implement navigation or LBS (Location Based Service). Positioning component 708 can be a positioning component based on the US GPS (Global Positioning System), China's Beidou system or Russia's Galileo system.

[0214] The power supply 709 is used to power various components in the terminal 700. The power supply 709 can be an alternating current, a direct current, a disposable battery, or a rechargeable battery. When the power supply 709 includes a rechargeable battery, the rechargeable battery can be a wired rechargeable battery or a wireless rechargeable battery. A wired rechargeable battery is a battery charged through a wired line, and a wireless rechargeable battery is a battery charged through a wireless coil. The rechargeable battery can also be used to support fast charging technology.

[0215] In some embodiments, the terminal 700 further includes one or more sensors 710 , including but not limited to: an acceleration sensor 711 , a gyroscope sensor 712 , a pressure sensor 713 , a fingerprint sensor 714 , an optical sensor 715 , and a proximity sensor 716 .

[0216] The acceleration sensor 711 can detect the magnitude of acceleration on the three coordinate axes of the coordinate system established by the terminal 700. For example, the acceleration sensor 711 can be used to detect the components of gravity acceleration on the three coordinate axes. The processor 701 can control the display screen 705 to display the user interface in a horizontal view or a vertical view according to the gravity acceleration signal collected by the acceleration sensor 711. The acceleration sensor 711 can also be used to collect motion data of games or users.

[0217] The gyro sensor 712 can detect the body direction and rotation angle of the terminal 700, and the gyro sensor 712 can cooperate with the acceleration sensor 711 to collect the user's 3D actions on the terminal 700. The processor 701 can implement the following functions based on the data collected by the gyro sensor 712: motion sensing (such as changing the UI according to the user's tilt operation), image stabilization during shooting, game control, and inertial navigation.

[0218] The pressure sensor 713 can be set on the side frame of the terminal 700 and / or the lower layer of the display screen 705. When the pressure sensor 713 is set on the side frame of the terminal 700, it can detect the user's holding signal of the terminal 700, and the processor 701 performs left and right hand recognition or shortcut operation according to the holding signal collected by the pressure sensor 713. When the pressure sensor 713 is set on the lower layer of the display screen 705, the processor 701 controls the operability controls on the UI interface according to the user's pressure operation on the display screen 705. The operability controls include at least one of a button control, a scroll bar control, an icon control, and a menu control.

[0219] The fingerprint sensor 714 is used to collect the user's fingerprint, and the processor 701 identifies the user's identity based on the fingerprint collected by the fingerprint sensor 714, or the fingerprint sensor 714 identifies the user's identity based on the collected fingerprint. When the user's identity is identified as a trusted identity, the processor 701 authorizes the user to perform relevant sensitive operations, including unlocking the screen, viewing encrypted information, downloading software, paying, and changing settings. The fingerprint sensor 714 can be set on the front, back, or side of the terminal 700. When a physical button or a manufacturer logo is set on the terminal 700, the fingerprint sensor 714 can be integrated with the physical button or the manufacturer logo.

[0220] The optical sensor 715 is used to collect the ambient light intensity. In one embodiment, the processor 701 can control the display brightness of the display screen 705 according to the ambient light intensity collected by the optical sensor 715. Specifically, when the ambient light intensity is high, the display brightness of the display screen 705 is increased; when the ambient light intensity is low, the display brightness of the display screen 705 is reduced. In another embodiment, the processor 701 can also dynamically adjust the shooting parameters of the camera assembly 706 according to the ambient light intensity collected by the optical sensor 715.

[0221] The proximity sensor 716, also called a distance sensor, is usually arranged on the front panel of the terminal 700. The proximity sensor 716 is used to collect the distance between the user and the front of the terminal 700. In one embodiment, when the proximity sensor 716 detects that the distance between the user and the front of the terminal 700 is gradually decreasing, the processor 701 controls the display screen 705 to switch from the screen-on state to the screen-off state; when the proximity sensor 716 detects that the distance between the user and the front of the terminal 700 is gradually increasing, the processor 701 controls the display screen 705 to switch from the screen-off state to the screen-on state.

[0222] Those skilled in the art will understand that Figure 7 The structure shown in the figure does not constitute a limitation on the terminal 700, and the terminal 700 may include more or less components than those shown in the figure, or combine some components, or adopt a different component arrangement.

[0223] Figure 8 It is a block diagram of a server provided by an embodiment of the present disclosure. The server 800 may have relatively large differences due to different configurations or performances, and may include one or more processors (Central Processing Units, CPU) 801 and one or more memories 802, wherein the memory 802 is used to store executable instructions, and the processor 801 is configured to execute the above executable instructions to implement the voice data processing method provided by the above-mentioned various method embodiments. Of course, the server may also have components such as a wired or wireless network interface, a keyboard, and an input and output interface for input and output. The server may also include other components for implementing device functions, which will not be described in detail here.

[0224] In an exemplary embodiment, a storage medium including instructions is also provided, such as a memory 802 including instructions, and the instructions can be executed by a processor 801 of a server 800 to complete the above-mentioned voice data processing method. Optionally, the storage medium can be a non-temporary computer-readable storage medium, for example, the non-temporary computer-readable storage medium can be a ROM (Read-Only Memory), a RAM (Random Access Memory), a CD-ROM (Compact Disc Read-Only Memory), a magnetic tape, a floppy disk, an optical data storage device, etc.

[0225] An embodiment of the present application also provides a computer-readable storage medium, in which at least one instruction is stored, and the at least one instruction is loaded and executed by a processor to implement the method for processing voice data of any of the above implementation modes.

[0226] An embodiment of the present application also provides a computer program product, which includes at least one program code, and the at least one program code is loaded and executed by a processor to implement the voice data processing method of any of the above implementation modes.

[0227] In some embodiments, the computer program involved in the embodiments of the present application may be deployed and executed on one electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed at multiple locations and interconnected by a communication network. Multiple electronic devices distributed at multiple locations and interconnected by a communication network may constitute a blockchain system.

[0228] An embodiment of the present application provides a method for processing voice data. Since the method can determine a target voice data set corresponding to the population characteristics in multiple voice data sets based on the population characteristics corresponding to the voice data to be processed, the comparison range of the voice data is narrowed; and then the target voice data matching the voice data is determined from the target voice data set, which reduces the difficulty of comparison, reduces the time spent on comparison, and thus improves the processing efficiency of the voice data.

[0229] The above are only optional embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for processing speech data, characterized in that: The method comprises: Determining voice data, where the voice data is voice data to be processed; The voice data is input into a voiceprint recognition model, and based on a plurality of classification task modules, a plurality of crowd features corresponding to the voice data are determined, and the plurality of crowd features are output, wherein the number of the crowd features is multiple, and the voiceprint recognition model includes a plurality of classification task modules, and each classification task module is used to obtain a crowd feature; the crowd feature is used to indicate the crowd to which the object of the voice data belongs; Determining a target voice data set corresponding to the crowd characteristics from a plurality of voice data sets, wherein the plurality of voice data sets are obtained based on the crowd characteristics, and each voice data set corresponds to a different crowd characteristic; Determining, from the target voice data set, target voice data that matches the voice data, wherein the voiceprint features of the target voice data match the voiceprint features of the voice data; The training process of the voiceprint recognition model includes: Input the sample voice data into the initial voiceprint recognition model, and output the prediction result of each classification task module for the sample voice data; for each classification task module, compare the prediction result with the marking result of the sample voice data to obtain a comparison result; determine the first loss value corresponding to the comparison result; based on the first loss values ​​of the multiple classification task modules, adjust the model parameters of each classification task module respectively; determine the corrected second loss value of each classification task module and the weight of each classification task module at the current moment; weightedly sum the second loss values ​​and weights of the multiple classification task modules to obtain the comprehensive loss value of the voiceprint recognition model; based on the comprehensive loss value, adjust the model parameters of each classification task module again to obtain the voiceprint recognition model.

2. The method according to claim 1, characterized in that The multiple population characteristics include gender and age, the multiple classification task modules include a gender classification module and an age classification module, and the inputting of the voice data into the voiceprint recognition model and determining the multiple population characteristics corresponding to the voice data based on the multiple classification task modules include: The voice data is input into the voiceprint recognition model, and the gender and age corresponding to the voice data are determined based on the gender classification module and the age classification module respectively.

3. The method according to claim 1, characterized in that The process of determining the weight of each classification task module at the current moment includes: For each classification task module, determining a weight adjustment parameter of the classification task module and a measurement index value of the classification task module at the current moment; Based on the measurement index value, determining a target logarithmic value, wherein the target logarithmic value is logarithmically related to the measurement index value; Based on the measurement index value, determining a target difference value, wherein the target difference value is negatively correlated with the measurement index value; Determine an exponential value with the target difference as a base and the weight adjustment parameter as an exponent; The negative of the product of the index value and the target logarithm value is determined to obtain the weight of the classification task module.

4. The method according to claim 1, characterized in that: The voiceprint recognition model includes a voiceprint classification module, and the voiceprint classification module is used to determine target voice data that matches the voice data. The determining the target voice data that matches the voice data from the target voice data set includes: The voice data is input into the voiceprint recognition model, and based on the voiceprint classification module, the voiceprint features of the voice data are determined, and based on the voiceprint features, the target voice data is determined from the target voice data set.

5. The method according to claim 1, characterized in that There are multiple crowd features, and determining a target voice data set corresponding to the crowd features from multiple voice data sets includes: Determine a speech data set corresponding to each crowd feature respectively, wherein the speech data set includes at least one speech data corresponding to the crowd feature; The intersection of the speech data sets corresponding to the plurality of population characteristics is obtained to obtain the target speech data set.

6. A speech data processing device, characterized in that: The device comprises: A first determining module, used to determine voice data, where the voice data is voice data to be processed; A second determination module is used to input the voice data into a voiceprint recognition model, determine a plurality of population features corresponding to the voice data based on a plurality of classification task modules, and output the plurality of population features, wherein the number of the population features is multiple, the voiceprint recognition model includes a plurality of classification task modules, and each classification task module is used to obtain a population feature; the population feature is used to indicate the population to which the object of the voice data belongs; A third determination module is used to determine a target voice data set corresponding to the crowd feature from a plurality of voice data sets, wherein the plurality of voice data sets are obtained based on the crowd features, and each voice data set corresponds to a different crowd feature; A fourth determination module is used to determine target voice data matching the voice data from the target voice data set, wherein the voiceprint feature of the target voice data matches the voiceprint feature of the voice data; The device also includes: An input-output module, used to input the sample voice data into the initial voiceprint recognition model, and output the prediction result of each classification task module on the sample voice data; A comparison module, used for comparing the prediction result with the labeling result of the sample speech data for each classification task module to obtain a comparison result; A fifth determining module, used to determine a first loss value corresponding to the comparison result; A first adjustment unit, configured to adjust the model parameters of each classification task module based on the first loss values ​​of the multiple classification task modules; A first determining unit, used to determine the corrected second loss value of each classification task module and the weight of each classification task module at a current moment; A second determining unit, configured to perform a weighted summation of the second loss values ​​and weights of the plurality of classification task modules to obtain a comprehensive loss value of the voiceprint recognition model; The second adjustment unit is used to adjust the model parameters of each classification task module again based on the comprehensive loss value to obtain the voiceprint recognition model.

7. An electronic device, characterized in that: The electronic device includes one or more processors and one or more memories, wherein the one or more memories store at least one instruction, and the at least one instruction is loaded and executed by the one or more processors to implement the voice data processing method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement the method for processing voice data according to any one of claims 1 to 5.

9. A computer program product, characterized in that The computer program product includes at least one program code, and the at least one program code is loaded and executed by a processor to implement the method for processing speech data according to any one of claims 1 to 5.

Images