Voiceprint recognition method and device based on artificial intelligence, computer equipment and medium

Abstract

The invention belongs to the field of artificial intelligence and the field of financial science and technology, and relates to a voiceprint recognition method and device based on artificial intelligence, computer equipment and a storage medium. Performing emotion recognition on the telephone voice to obtain an emotion recognition result, and performing emotion labeling on the text based on the emotion recognition result to obtain a target text; calling a service recognition model to recognize the target text to obtain service scene information; obtaining a target audio preprocessing strategy corresponding to the service scene information, and preprocessing the telephone recording based on the target audio preprocessing strategy to obtain target voice data; screening the plurality of voiceprint recognition models based on the screening strategy and the service scene information to determine a target voiceprint recognition model; and performing voiceprint recognition on the target voice data based on the target voiceprint recognition model to obtain a voiceprint recognition result. According to the method and the device, the determination accuracy of the voiceprint recognition model is improved, and the voiceprint recognition accuracy of the telephone recording is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology and the field of financial technology, and in particular to artificial intelligence-based voiceprint recognition methods, devices, computer equipment and storage media. Background Art

[0002] With the vigorous development of the financial technology field and the accelerated advancement of digital transformation, voiceprint recognition technology, as an important branch of biometric technology, has shown great application potential and value in the financial industry. This technology achieves efficient and contactless identity authentication by analyzing and identifying an individual's unique voice characteristics, namely voiceprints, and provides a strong guarantee for the security and convenience of financial transactions. Voiceprint recognition technology plays an indispensable role in many key links of financial services, such as customer identity authentication, fraud detection and risk management.

[0003] However, despite the increasing application of voiceprint recognition technology in the financial industry, the current common practice is to deploy a single universal voiceprint model on the server side. When trying to cover all business scenarios, a single voiceprint model often finds it difficult to accurately capture and distinguish these subtle but critical differences, resulting in a decrease in recognition accuracy, which in turn affects the reliability of identity authentication and the sensitivity of fraud detection. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose a voiceprint recognition method, device, computer equipment and storage medium based on artificial intelligence to solve the technical problem of reduced recognition accuracy in the practice of deploying a single universal voiceprint model on the server side adopted by the existing financial industry.

[0005] In order to solve the above technical problems, the embodiment of the present application provides a voiceprint recognition method based on artificial intelligence, which adopts the following technical solutions:

[0006] Obtaining a telephone recording to be processed, and transcribing the telephone recording to obtain a corresponding text;

[0007] Performing emotion recognition on the telephone recording to obtain a corresponding emotion recognition result, and performing emotion annotation on the text based on the emotion recognition result to obtain a corresponding target text;

[0008] Calling a pre-trained business identification model; wherein the business identification model is a model obtained by training a pre-trained language model based on pre-collected telephone recording sample data;

[0009] Performing recognition processing on the target text based on the business recognition model to obtain business scenario information corresponding to the target text;

[0010] Acquire a target audio preprocessing strategy corresponding to the business scenario information, and preprocess the telephone recording based on the target audio preprocessing strategy to obtain corresponding target voice data;

[0011] Based on the preset screening strategy and the business scenario information, the preset multiple voiceprint recognition models are screened and processed to determine the corresponding target voiceprint recognition model;

[0012] The target voiceprint recognition model is used to perform voiceprint recognition processing on the target speech data to obtain a corresponding voiceprint recognition result.

[0013] Furthermore, the step of screening the preset multiple voiceprint recognition models based on the preset screening strategy and the business scenario information to determine the corresponding target voiceprint recognition model specifically includes:

[0014] Acquire all the voiceprint recognition models, and select a first voiceprint recognition model that meets the business scenario information from all the voiceprint recognition models;

[0015] Collecting performance data, resource consumption data, stability data, maintainability data, and cost data of each of the first voiceprint recognition models based on the target source;

[0016] Generate, based on a preset scoring strategy, a performance score, a resource consumption score, a stability score, a maintainability score, and a cost score for each of the first voiceprint recognition models corresponding to the performance data, resource consumption data, stability data, maintainability data, and cost data of each of the first voiceprint recognition models;

[0017] generating a model score for each of the first voiceprint recognition models based on the performance score, resource consumption score, stability score, maintainability score, and cost score of each of the first voiceprint recognition models;

[0018] Obtain the target model score with the highest value from all the model scores;

[0019] A second voiceprint recognition model corresponding to the target model score is screened out from all the first voiceprint recognition models, and the second voiceprint recognition model is used as the target voiceprint recognition model.

[0020] Furthermore, the step of generating a model score of each of the first voiceprint recognition models based on the performance score, resource consumption score, stability score, maintainability score and cost score of each of the first voiceprint recognition models specifically includes:

[0021] Get the preset weight setting algorithm;

[0022] Generating factor weights corresponding to the performance factor, the resource consumption factor, the stability factor, the maintainability factor and the cost factor respectively based on the weight setting algorithm;

[0023] Call the preset score calculation formula;

[0024] Based on the score calculation formula, the performance score, resource consumption score, stability score, maintainability score, cost score and the factor weight of each of the first voiceprint recognition models are calculated and processed to obtain a model score of each of the first voiceprint recognition models.

[0025] Furthermore, the step of obtaining a target audio preprocessing strategy corresponding to the business scenario information specifically includes:

[0026] Call the preset preprocessing strategy library;

[0027] Performing matching processing on the preprocessing policy library based on the business scenario information, so as to query a specified audio preprocessing policy matching the business scenario information from the preprocessing policy library;

[0028] The specified audio preprocessing strategy is used as the target audio preprocessing strategy.

[0029] Furthermore, the step of performing emotion recognition on the telephone recording to obtain a corresponding emotion recognition result, and performing emotion annotation on the text based on the emotion recognition result to obtain a corresponding target text specifically includes:

[0030] Call the preset emotion recognition model;

[0031] Inputting the telephone recording into the emotion recognition model, performing emotion recognition processing on the telephone recording through the emotion recognition model and outputting corresponding emotion recognition results;

[0032] Performing emotion annotation on the text based on the emotion recognition result to obtain annotated text;

[0033] The annotated text is used as the target text.

[0034] Furthermore, before the step of calling the pre-trained service identification model, the method further includes:

[0035] Acquire a pre-collected telephone recording data set; wherein the telephone recording data set includes designated telephone recordings in various business scenarios and labels corresponding to the designated telephone recordings;

[0036] Preprocessing the telephone recording data set to obtain corresponding telephone recording sample data;

[0037] Call the pre-trained language model;

[0038] Based on the telephone recording sample data and the preset low-rank adapter, the pre-trained language model is trained and fine-tuned to obtain a specified language model that meets the expected conditions;

[0039] The specified language model is used as the service identification model.

[0040] Furthermore, after the step of performing voiceprint recognition processing on the target voice data based on the target voiceprint recognition model to obtain a corresponding voiceprint recognition result, the method further includes:

[0041] Get the preset formatting strategy;

[0042] Formatting the voiceprint recognition result based on the formatting strategy to obtain a corresponding target voiceprint recognition result;

[0043] Generating explanation information corresponding to the voiceprint recognition result;

[0044] Determine the target presentation method;

[0045] Based on the target display method, the target voiceprint recognition result and the interpretation information are sent to a preset downstream system.

[0046] In order to solve the above technical problems, the embodiment of the present application also provides a voiceprint recognition device based on artificial intelligence, which adopts the following technical solution:

[0047] A transcription module, used to obtain the telephone recording to be processed, and transcribe the telephone recording to obtain the corresponding text;

[0048] A tagging module, used for performing emotion recognition on the telephone recording to obtain a corresponding emotion recognition result, and performing emotion tagging on the text based on the emotion recognition result to obtain a corresponding target text;

[0049] A first calling module is used to call a pre-trained service identification model; wherein the service identification model is a model obtained by training a pre-trained language model based on pre-collected telephone recording sample data;

[0050] A first recognition module, configured to perform recognition processing on the target text based on the business recognition model to obtain business scenario information corresponding to the target text;

[0051] A first acquisition module is used to acquire a target audio preprocessing strategy corresponding to the business scenario information, and preprocess the telephone recording based on the target audio preprocessing strategy to obtain corresponding target voice data;

[0052] A screening module, used to screen the preset multiple voiceprint recognition models based on the preset screening strategy and the business scenario information, and determine the corresponding target voiceprint recognition model;

[0053] The second recognition module is used to perform voiceprint recognition processing on the target speech data based on the target voiceprint recognition model to obtain a corresponding voiceprint recognition result.

[0054] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0055] The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any one of the artificial intelligence-based voiceprint recognition methods proposed in the embodiments of the present application are implemented.

[0056] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0057] The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the steps of any one of the artificial intelligence-based voiceprint recognition methods proposed in the embodiments of the present application are implemented.

[0058] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0059] The present application obtains the corresponding text by transcribing the acquired telephone recording, obtains the emotion recognition result by performing emotion recognition on the telephone voice, and obtains the corresponding target text by performing emotion annotation on the text based on the emotion recognition result, and then obtains the corresponding target text by using the business recognition model to obtain the corresponding business scenario information, and then obtains the target audio preprocessing strategy corresponding to the business scenario information, and preprocesses the telephone recording based on the target audio preprocessing strategy to obtain the corresponding target voice data, and then intelligently screens the preset multiple voiceprint recognition models based on the preset screening strategy and the business scenario information to determine the corresponding target voiceprint recognition model, and the screened target voiceprint recognition model is a model adapted to the business scenario information, which effectively improves the determination accuracy of the target voiceprint recognition model, and improves the flexibility and reliability of voiceprint recognition model screening, which is beneficial to the subsequent use of the target voiceprint recognition model to perform voiceprint recognition processing on the target voice data, and can ensure that the performance and effect of the target voiceprint recognition model in the voiceprint recognition processing process meet expectations, thereby facilitating the improvement of the recognition accuracy of voiceprint recognition for telephone recordings. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0061] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0062] Figure 2 A flowchart of an embodiment of a voiceprint recognition method based on artificial intelligence according to the present application;

[0063] Figure 3 is a structural schematic diagram of an embodiment of a voiceprint recognition device based on artificial intelligence according to the present application;

[0064] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0065] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0066] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0067] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0068] like Figure 1As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0069] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0070] The terminal device 101 can be various electronic devices with a display screen and supporting web browsing. In addition to the laptop computer 1011, the tablet computer 1012 or the mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV) player, a laptop computer, a desktop computer, etc.

[0071] The server 103 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal device 101 .

[0072] It should be noted that the artificial intelligence-based voiceprint recognition method provided in the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based voiceprint recognition device is generally arranged in a server / terminal device.

[0073] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0074] Continue to refer Figure 2, showing a flow chart of an embodiment of a voiceprint recognition method based on artificial intelligence according to the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The voiceprint recognition method based on artificial intelligence provided in the embodiment of the present application can be applied to any scenario that requires voiceprint recognition, and the voiceprint recognition method based on artificial intelligence can be applied to products in these scenarios, for example, voiceprint recognition in insurance call scenarios in the financial insurance field. The voiceprint recognition method based on artificial intelligence includes the following steps:

[0075] Step S201, obtaining a telephone recording to be processed, and transcribing the telephone recording to obtain a corresponding text.

[0076] In this embodiment, the voiceprint recognition method based on artificial intelligence is run on the electronic device (for example Figure 1 The server / terminal device shown in the figure) can obtain the telephone recording to be processed through a wired connection or a wireless connection. It should be noted that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, Wi F i connection, Bluetooth connection, Wi MAX connection, Zigbee connection, UWB (ultrawi deband) connection, and other wireless connection methods currently known or to be developed in the future. The present application can be applied to fields such as customer identity authentication, anti-fraud and risk management in the financial industry. The above-mentioned telephone recording to be processed may refer to telephone recordings in the process of business processing in various business departments of a financial enterprise, such as customer service centers, telemarketing centers, claims centers, etc. Among them, the telephone recordings can be converted into text by using automatic speech recognition (ASR) technology. The transcription process ensures that the data processed subsequently is structured text data and is convenient for further analysis.

[0077] Step S202, performing emotion recognition on the telephone recording to obtain a corresponding emotion recognition result, and performing emotion annotation on the text based on the emotion recognition result to obtain a corresponding target text.

[0078] In this embodiment, the specific implementation process of performing emotion recognition on the telephone recording to obtain the corresponding emotion recognition result, and performing emotion annotation on the text based on the emotion recognition result to obtain the corresponding target text will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0079] Step S203, calling a pre-trained business identification model; wherein the business identification model is a model obtained by training a pre-trained language model based on pre-collected telephone recording sample data.

[0080] In this embodiment, the construction process of the above-mentioned business identification model will be further described in detail in subsequent specific embodiments of the present application, and will not be elaborated in detail here.

[0081] Step S204: performing recognition processing on the target text based on the business recognition model to obtain business scenario information corresponding to the target text.

[0082] In this embodiment, after the target text is input into the business identification model, the business identification model will perform reasoning processing based on the target text, calculate through forward propagation, and output the judgment result of the corresponding business scenario. Among them, the business identification model can be set to reasoning mode, and unnecessary calculations (such as dropout) can be turned off to speed up the reasoning process. Specifically, during the reasoning process, the business identification model will identify key information in the target text, such as the participants in the conversation, the topics mentioned, the language style used, etc. Based on this information, the business identification model can classify and label the calls and determine their business scenario information. The business scenario information corresponds to the call type or attribute, which may specifically include outbound calls, inbound calls, business consultations, claims processing, etc.

[0083] Step S205, obtaining a target audio preprocessing strategy corresponding to the business scenario information, and preprocessing the telephone recording based on the target audio preprocessing strategy to obtain corresponding target voice data.

[0084] In this embodiment, the specific implementation process of obtaining the target audio preprocessing strategy corresponding to the business scenario information will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated on here. After selecting the target audio preprocessing strategy, the preprocessing module corresponding to the target audio preprocessing strategy can be loaded, and the telephone recording can be input into the preprocessing module to perform denoising, voice extraction, echo cancellation and other operations, so as to obtain the corresponding target voice data.

[0085] Step S206, screening the preset multiple voiceprint recognition models based on the preset screening strategy and the business scenario information to determine the corresponding target voiceprint recognition model.

[0086] In this embodiment, the above-mentioned specific implementation process of screening and processing the preset multiple voiceprint recognition models based on the preset screening strategy and the business scenario information to determine the corresponding target voiceprint recognition model will be further described in detail in the subsequent specific embodiments of this application and will not be elaborated in detail here.

[0087] Step S207: performing voiceprint recognition processing on the target speech data based on the target voiceprint recognition model to obtain a corresponding voiceprint recognition result.

[0088] In this embodiment, the target voice data is input into the target voiceprint recognition model, and the corresponding voiceprint recognition algorithm is executed, for example, voiceprint recognition processing steps including feature extraction, voiceprint comparison, matching degree calculation, etc., to obtain the corresponding voiceprint recognition result.

[0089] The present application obtains the corresponding text by transcribing the acquired telephone recording, obtains the emotion recognition result by performing emotion recognition on the telephone voice, and obtains the corresponding target text by performing emotion annotation on the text based on the emotion recognition result, and then obtains the corresponding target text by using the business recognition model to obtain the corresponding business scenario information, and then obtains the target audio preprocessing strategy corresponding to the business scenario information, and preprocesses the telephone recording based on the target audio preprocessing strategy to obtain the corresponding target voice data, and then intelligently screens the preset multiple voiceprint recognition models based on the preset screening strategy and the business scenario information to determine the corresponding target voiceprint recognition model, and the screened target voiceprint recognition model is a model adapted to the business scenario information, which effectively improves the determination accuracy of the target voiceprint recognition model, and improves the flexibility and reliability of voiceprint recognition model screening, which is beneficial to the subsequent use of the target voiceprint recognition model to perform voiceprint recognition processing on the target voice data, and can ensure that the performance and effect of the target voiceprint recognition model in the voiceprint recognition processing process meet expectations, thereby facilitating the improvement of the recognition accuracy of voiceprint recognition for telephone recordings.

[0090] In some optional implementations, step S206 includes the following steps:

[0091] All the voiceprint recognition models are obtained, and a first voiceprint recognition model that meets the business scenario information is screened out from all the voiceprint recognition models.

[0092] In this embodiment, a database containing multiple voiceprint recognition models is pre-established. These voiceprint recognition models may be optimized for different call types (such as outbound calls, incoming calls), audio preprocessing methods (such as denoising, human voice enhancement), and different audio characteristics (such as sound quality, volume, speaking speed, etc.). All the voiceprint recognition models can be called from the above database. The technical documents of all voiceprint recognition models can be consulted, and the applicable scenario information of each voiceprint recognition model can be determined according to the technical documents. Then, according to the applicable scenario information, the first voiceprint recognition model that meets the business scenario information can be screened out from all the voiceprint recognition models. Among them, meeting the business scenario information means that the rating of the applicable scenario information corresponding to the business scenario information in the voiceprint recognition model is medium or high.

[0093] Performance data, resource consumption data, stability data, maintainability data and cost data of each of the first voiceprint recognition models are collected based on the target source.

[0094] In this embodiment, the target sources may at least include the model's technical documentation, test reports, user feedback and other sources. The performance data may refer to the recognition accuracy of the model, the resource consumption data may refer to the CPU usage or response time, the stability data may refer to the stability of the model when running for a long time or processing a large amount of data, the maintainability data may refer to the model's documentation quality, community support, update frequency, etc., and the cost data may refer to the cost of purchasing, deploying, and maintaining the model.

[0095] Based on a preset scoring strategy, a performance score, a resource consumption score, a stability score, a maintainability score and a cost score of each of the first voiceprint recognition models corresponding to the performance data, resource consumption data, stability data, maintainability data and cost data of each of the first voiceprint recognition models are generated.

[0096] In this embodiment, according to the actual model evaluation requirements, corresponding scoring strategies are set in advance for performance factors, resource consumption factors, stability factors, maintainability factors and cost factors, and the scoring strategies are strategies based on scoring criteria. The scoring criteria can be quantitative or qualitative, depending on the nature of the factors. The scoring criteria include quantitative scoring criteria and qualitative scoring criteria. Specifically, quantitative scoring criteria: for performance factors, specific numerical values ​​(such as 5 points for 95% accuracy, 3 points for 85%, etc.) or percentage ranges can be directly used as scoring criteria. For resource consumption factors, resource usage (such as 5 points for CPU usage less than 5%, 4 points for 10%-15%, etc.) or response time (such as 5 points for less than 100 milliseconds, 4 points for 100-200 milliseconds, etc.) can be used as scoring criteria. Qualitative scoring criteria: For factors that are difficult to quantify, such as stability factors, maintainability factors and cost factors, qualitative scoring criteria are set in advance. For example, stability can be divided into four levels: "very stable", "stable", "average", and "unstable", and different scores can be assigned to each level: "very stable" corresponds to 5 points, "stable" corresponds to 4 points, "average" corresponds to 3 points, and "unstable" corresponds to 2 points. Maintainability factors can be divided into three levels: "good", "average", and "poor", and different scores can be assigned to each level: "good" corresponds to 5 points, "average" corresponds to 3 points, and "poor" corresponds to 1 point. Cost factors can be divided into three levels: "greater than 2k", "1-2k", and "less than 1k", and different scores can be assigned to each level: "greater than 2k" corresponds to 1 point, "1-2k" corresponds to 3 points, and "less than 1k" corresponds to 5 points.

[0097] A model score of each of the first voiceprint recognition models is generated based on the performance score, resource consumption score, stability score, maintainability score and cost score of each of the first voiceprint recognition models.

[0098] In this embodiment, the specific implementation process of generating the model score of each of the first voiceprint recognition models based on the performance score, resource consumption score, stability score, maintainability score and cost score of each of the first voiceprint recognition models will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0099] The target model score with the highest value is obtained from all the model scores.

[0100] In this embodiment, numerical comparison may be performed on all the model scores to obtain the target model score with the highest numerical value from all the model scores.

[0101] A second voiceprint recognition model corresponding to the target model score is screened out from all the first voiceprint recognition models, and the second voiceprint recognition model is used as the target voiceprint recognition model.

[0102] In this embodiment, the target voiceprint recognition model is a model that matches the target model score among all the first voiceprint recognition models.

[0103] The present application first selects a first voiceprint recognition model that meets the business scenario information from all the voiceprint recognition models, and then generates, based on a preset scoring strategy, a performance score, a resource consumption score, a stability score, a maintainability score and a cost score of each of the first voiceprint recognition models corresponding to the performance data, resource consumption data, stability data, maintainability data and cost data of each of the first voiceprint recognition models collected from a target source, and then generates a model score of each of the first voiceprint recognition models based on the performance score, resource consumption score, stability score, maintainability score and cost score of each of the first voiceprint recognition models, and then selects a first voiceprint recognition model that meets the business scenario information from all the first voiceprint recognition models. The second voiceprint recognition model corresponding to the target model score with the highest numerical value is screened out from the voiceprint recognition model and used as the final target voiceprint recognition model. Since the target voiceprint recognition model conforms to the above-mentioned business scenario information and has the optimal comprehensive capabilities corresponding to the performance factor, resource consumption factor, stability factor, maintainability factor and cost factor, the determination accuracy of the target voiceprint recognition model is effectively improved, and the flexibility and reliability of voiceprint recognition model screening are improved, which is conducive to the subsequent use of the target voiceprint recognition model to perform voiceprint recognition processing on the target voice data, and can ensure that the performance and effect of the target voiceprint recognition model in the processing process meet expectations.

[0104] In some optional implementations of this embodiment, generating the model score of each of the first voiceprint recognition models based on the performance score, resource consumption score, stability score, maintainability score, and cost score of each of the first voiceprint recognition models comprises the following steps:

[0105] Get the preset weight setting algorithm.

[0106] In this embodiment, there is no specific limitation on the selection of the above-mentioned weight setting algorithm, which can be determined according to actual business processing needs. Specifically, any one of the standardization method, AHP hierarchical analysis method, principal component analysis method, factor analysis method, entropy method, etc. can be adopted.

[0107] Factor weights corresponding to the performance factor, resource consumption factor, stability factor, maintainability factor and cost factor are generated based on the weight setting algorithm.

[0108] In this embodiment, the performance factor, resource consumption factor, stability factor, maintainability factor and cost factor may be weighted according to the corresponding weight allocation processing steps in the weight setting algorithm, thereby generating factor weights corresponding to the performance factor, resource consumption factor, stability factor, maintainability factor and cost factor, respectively. The factor weights may include a first weight, a second weight, a third weight, a fourth weight and a fifth weight corresponding to the performance factor, resource consumption factor, stability factor, maintainability factor and cost factor, respectively.

[0109] Call the preset score calculation formula.

[0110] In this embodiment, the score calculation formula may specifically adopt a weighted sum formula.

[0111] Based on the score calculation formula, the performance score, resource consumption score, stability score, maintainability score, cost score and the factor weight of each of the first voiceprint recognition models are calculated and processed to obtain a model score of each of the first voiceprint recognition models.

[0112] In this embodiment, the model score of each of the first voiceprint recognition models can be obtained by substituting the performance score, resource consumption score, stability score, maintainability score, cost score and the factor weights of the first voiceprint recognition model into the corresponding positions in the above score calculation formula.

[0113] The present application generates factor weights corresponding to performance factors, resource consumption factors, stability factors, maintainability factors and cost factors respectively based on a preset weight setting algorithm, and then uses the performance scores, resource consumption scores, stability scores, maintainability scores, cost scores and the factor weights of each first voiceprint recognition model based on the score calculation formula for calculation processing, so as to achieve rapid and accurate generation of model scores for each first voiceprint recognition model, and effectively ensure the accuracy of the generated model scores. This is conducive to subsequently screening out the model corresponding to the target model score with the highest numerical value according to the model scores of each first voiceprint recognition model obtained and using it as the final target voiceprint recognition model, thereby improving the accuracy of the generated target voiceprint recognition model.

[0114] In some optional implementations, the obtaining of the target audio preprocessing strategy corresponding to the business scenario information in step S205 includes the following steps:

[0115] Call the preset preprocessing strategy library.

[0116] In this embodiment, according to different types of business scenarios (such as different call types or attributes), a set of preprocessing strategy libraries for audio processing are predefined. The preprocessing strategy library contains multiple preprocessing strategies, and each preprocessing strategy is optimized for a specific business scenario. Among them, the above-mentioned business scenarios may at least include outbound calls, inbound calls, business consultations, and claims processing. Correspondingly, 1) the preprocessing strategy corresponding to the business scenario of outbound calls is: efficient denoising, voice extraction, and the remark information is: for shorter customer voices, a fast algorithm is used. 2) The preprocessing strategy corresponding to the business scenario of inbound calls is: detailed denoising, echo cancellation, voice enhancement, and the remark information is: for longer customer voices, ensure clear sound quality. 3) The preprocessing strategy corresponding to the business scenario of business consultation is: keyword extraction, voice segmentation, and the remark information is: for subsequent analysis. 4) The preprocessing strategy corresponding to the business scenario of claims processing is: complete recording retention, emotion annotation, and the remark information is: it may be necessary to analyze the emotional information in the voice.

[0117] The pre-processing policy library is matched based on the business scenario information to query the pre-processing policy library for a designated audio pre-processing policy that matches the business scenario information.

[0118] In this embodiment, the preprocessing strategy library can be matched by using the business scenario information, first determining the target business scenario information that is the same as the above business scenario information, and then querying the audio preprocessing strategy that matches the above target business scenario information from the preprocessing strategy library to obtain the above specified audio preprocessing strategy. Exemplarily, if the business identification model determines that the current call corresponding to the telephone recording is an outbound call, the audio preprocessing strategy corresponding to the efficient denoising and voice extraction algorithm is selected as the matching specified audio preprocessing strategy and used for subsequent preprocessing.

[0119] The specified audio preprocessing strategy is used as the target audio preprocessing strategy.

[0120] In this embodiment, the parameters of the selected audio preprocessing strategy can be adjusted according to actual needs, for example, the threshold of the denoising algorithm, the sensitivity of the voice extraction algorithm, etc. can be adjusted to ensure the best preprocessing effect.

[0121] The present application uses a preset preprocessing strategy library and matches the preprocessing strategy library based on the business scenario information, so that the target audio preprocessing strategy that matches the business scenario information can be quickly and accurately queried from the preprocessing strategy library, so as to automatically and intelligently select the most suitable target audio preprocessing strategy corresponding to the telephone recording from the preprocessing strategy library, so that the target audio preprocessing strategy can be used to preprocess the telephone recording subsequently, which can effectively improve the efficiency and accuracy of audio processing and significantly reduce manual intervention and error rate.

[0122] In some optional implementations, the performing of emotion recognition on the telephone recording to obtain a corresponding emotion recognition result in step S202, and performing emotion annotation on the text based on the emotion recognition result to obtain a corresponding target text, includes the following steps:

[0123] Call the preset emotion recognition model.

[0124] In this embodiment, the emotion recognition model is pre-trained using speech data with emotion labels, and the model can be a rule-based method or a machine learning or deep learning model.

[0125] The telephone recording is input into the emotion recognition model, and the emotion recognition model is used to perform emotion recognition processing on the telephone recording and output a corresponding emotion recognition result.

[0126] In this embodiment, while the telephone recording is being transcribed, a trained emotion recognition model may be used to perform emotion recognition processing on the telephone recording and output a corresponding emotion recognition result.

[0127] The text is emotionally annotated based on the emotion recognition result to obtain an annotated text.

[0128] In this embodiment, according to the emotion recognition result, corresponding emotion keywords are added to the above text. These keywords can be predefined labels (such as "angry", "happy", etc.), or quantified values ​​of emotion intensity, so as to complete the emotion annotation processing of the text and obtain the annotated text.

[0129] The annotated text is used as the target text.

[0130] The present application calls the emotion recognition model to perform emotion recognition processing on the telephone recording and outputs the corresponding emotion recognition results, and then performs emotion annotation on the text based on the emotion recognition results, so as to quickly and accurately complete the emotion annotation processing and obtain the corresponding target text, thereby ensuring the data accuracy of the target text and making it available for use in subsequent steps.

[0131] In some optional implementations of this embodiment, before step S203, the electronic device may further perform the following steps:

[0132] Get a pre-collected dataset of phone recordings.

[0133] In this embodiment, the telephone recording data set includes designated telephone recordings in various business scenarios and labels corresponding to the designated telephone recordings.

[0134] The telephone recording data set is preprocessed to obtain corresponding telephone recording sample data.

[0135] In this embodiment, the above preprocessing may include steps such as ASR transcription, emotion labeling (if necessary), and text cleaning. The process of ASR transcription and emotion labeling can refer to the process of transcribing the above telephone speech, and the process of emotion labeling the text corresponding to the telephone speech, which will not be described in detail here.

[0136] Call the pre-trained language model.

[0137] In this embodiment, there is no specific limitation on the selection of the pre-trained language model, for example, LLM may be used.

[0138] The pre-trained language model is trained and fine-tuned based on the telephone recording sample data and a preset low-rank adapter to obtain a specified language model that meets expected conditions.

[0139] In this embodiment, the above-mentioned low-rank adapter is specifically LoRA. The process of constructing a specified language model based on the telephone recording sample data and LoRA may include: Freezing the pre-trained weights: First, freeze the pre-trained weights of the pre-trained language model to keep the general knowledge learned on the large-scale data set unchanged. Add a LoRA layer: Add a LoRA layer to the Attention module of the pre-trained language model to learn the incremental weight ΔW for a specific business scenario. Train the LoRA layer: By training the LoRA layer (i.e., adjusting the parameters of B and A), the model can better adapt to the judgment of call business scenario information under different business scenarios. Reasoning stage: In the reasoning stage, the trained LoRA weights are directly loaded to generate the judgment results of the corresponding business scenarios.

[0140] Specifically, the process of freezing pre-trained weights includes: 1. Loading the pre-trained language model: First, load the pre-trained large language model (LLM) from the model repository or local file system. 2. Setting weight freeze: After the model is loaded, traverse all layers of the model and set the pre-trained weights to an untrainable state. This can be achieved by modifying the model's required_grad attribute to false. 3. Verifying the frozen state: Ensure that all pre-trained weights have been correctly frozen by printing the model parameters or checking the required_grad attribute.

[0141] The process of adding a LoRA layer includes: 1. Select the insertion point: insert the LoRA layer in the Attention module of the pre-trained language model. 2. Define the LoRA layer: According to the principle of LoRA, define the incremental weight layer (ie, the LoRA layer). This includes initializing the incremental weight matrix ΔW, which can be low-rank and is composed of matrices B and A in some way (such as B*A). 3. Integrate the LoRA layer: Integrate the LoRA layer into the pre-trained language model to ensure that it can be called and calculated correctly during inference and training. This usually involves modifying the forward propagation function of the model. 4. Initialize the LoRA parameters: Initialize the parameters of the LoRA layer (B and A). B is usually initialized to a zero matrix to ensure that no additional noise is generated at the beginning of training; A can be initialized using Gaussian initialization or other suitable initialization methods.

[0142] The process of training the LoRA layer includes: 1. Setting training parameters: configuring the parameters required in the training process, such as learning rate, batch size, training rounds, etc. 2. Training loop: executing the training loop, inputting the preprocessed telephone recording sample data into the model, calculating the loss through forward propagation, and then updating the parameters (B and A) of the LoRA layer through back propagation until the training stop condition is reached, and the trained model that meets the expectations is used as the business identification model. Among them, since the pre-trained weights are frozen, only the parameters of the LoRA layer will be updated. 3. Verification and adjustment: the performance of the model can be verified regularly during the training process, and the training parameters can be adjusted as needed.

[0143] The specified language model is used as the service identification model.

[0144] The present application obtains telephone recording sample data by preprocessing the acquired telephone recording data set, and then uses the telephone recording sample data and a preset low-rank adapter to train and fine-tune the pre-trained language model, so as to quickly and intelligently complete the construction of the required business identification model, improve the construction efficiency of the business identification model, and ensure the model effect of the obtained business identification model.

[0145] In some optional implementations of this embodiment, after step S207, the electronic device may further perform the following steps:

[0146] Get the preset formatting strategy.

[0147] In this embodiment, the content of the formatting strategy includes: Step 1: Determine the output format. First, it is necessary to determine the output format of the voiceprint recognition result. Common formats include JSON (JavaScript Object Notation) and XML (Extensible Markup Language). The specific format selection may depend on the needs of the downstream system or the user's preference. Step 2: Design the data structure. Design a suitable data structure according to the content of the voiceprint recognition result. Exemplarily, an output in JSON format may contain the following fields: result: the main part of the recognition result. voice_id: the unique identifier of the recognized voiceprint. match_score: the matching score, which indicates the similarity between the recognition result and a known voiceprint. metadata: other related metadata, such as recognition time, audio file ID, etc. Step 3: Assemble the data. Assemble the voiceprint recognition results according to the designed data structure. Specifically, extract the necessary information from the output of the recognition algorithm and fill it into the corresponding fields. Step 4: Format the output. Format the assembled data using the selected format (such as JSON). For JSON, convert the data into key-value pairs and use quotes around keys and string values, and appropriate indentation or spaces to improve readability.

[0148] The voiceprint recognition result is formatted based on the formatting strategy to obtain a corresponding target voiceprint recognition result.

[0149] In this embodiment, the voiceprint recognition result may be formatted according to the policy content of the above formatting policy to obtain the corresponding target voiceprint recognition result.

[0150] Generate interpretation information corresponding to the voiceprint recognition result.

[0151] In this embodiment, before sending the voiceprint recognition result, the voiceprint recognition result may also be analyzed, which may include checking the matching score, the recognized voiceprint information and other metadata to determine the validity and accuracy of the voiceprint recognition result. Then, based on the analysis result, the above voiceprint recognition result is explained or illustrated to obtain corresponding explanation information. The content of the explanation information can use non-technical language and remain clear and concise so that users or downstream systems can easily understand it.

[0152] Determine the target presentation method.

[0153] In this embodiment, there is no limitation on the determination method of the above-mentioned target display method, which can be set according to actual business needs. For example, if the interaction is carried out through a Web interface, the target voiceprint recognition result and explanation information can be embedded in an HTML page; if the data is provided through an API interface, the target voiceprint recognition result and explanation field can be included in the response body; if the description is carried out through a document or report, the target voiceprint recognition result and explanation information can be written into the document.

[0154] Based on the target display method, the target voiceprint recognition result and the interpretation information are sent to a preset downstream system.

[0155] In this embodiment, the target voiceprint recognition result and the explanation information may be presented to the downstream system in a presentation manner corresponding to the target presentation manner.

[0156] After the target voiceprint recognition model is used to perform voiceprint recognition processing on the target voice data to obtain the voiceprint recognition result, the present application will intelligently format the voiceprint recognition result based on a preset formatting strategy to obtain the corresponding target voiceprint recognition result, and will further generate explanation information corresponding to the voiceprint recognition result, and then send the target voiceprint recognition result and the explanation information to the preset downstream system based on the target display method, thereby ensuring the standardization of the sending of the target voiceprint recognition result and improving the intelligence of the sending of the target voiceprint recognition result. In addition, the use of the explanation information can effectively help the downstream system understand the voiceprint recognition result, thereby improving the user experience.

[0157] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0158] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned voiceprint recognition results, the above-mentioned voiceprint recognition results can also be stored in a node of a blockchain.

[0159] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0160] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0161] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0162] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0163] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0164] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a voiceprint recognition device based on artificial intelligence, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0165] like Figure 3As shown, the artificial intelligence-based voiceprint recognition device 300 described in this embodiment includes: a transcription module 301, a labeling module 302, a first calling module 303, a first recognition module 304, a first acquisition module 305, a screening module 306 and a second recognition module 307.

[0166] in:

[0167] The transcription module 301 is used to obtain the telephone recording to be processed and transcribe the telephone recording to obtain the corresponding text;

[0168] The tagging module 302 is used to perform emotion recognition on the telephone recording to obtain a corresponding emotion recognition result, and perform emotion tagging on the text based on the emotion recognition result to obtain a corresponding target text;

[0169] The first calling module 303 is used to call a pre-trained service identification model; wherein the service identification model is a model obtained by training a pre-trained language model based on pre-collected telephone recording sample data;

[0170] A first recognition module 304 is used to perform recognition processing on the target text based on the business recognition model to obtain business scenario information corresponding to the target text;

[0171] A first acquisition module 305 is used to acquire a target audio preprocessing strategy corresponding to the business scenario information, and preprocess the telephone recording based on the target audio preprocessing strategy to obtain corresponding target voice data;

[0172] A screening module 306 is used to screen the preset multiple voiceprint recognition models based on the preset screening strategy and the business scenario information to determine the corresponding target voiceprint recognition model;

[0173] The second recognition module 307 is used to perform voiceprint recognition processing on the target speech data based on the target voiceprint recognition model to obtain a corresponding voiceprint recognition result.

[0174] In some optional implementations of this embodiment, the screening module 306 includes:

[0175] A screening submodule, used to obtain all the voiceprint recognition models, and screen out a first voiceprint recognition model that meets the business scenario information from all the voiceprint recognition models;

[0176] A collection submodule, configured to collect performance data, resource consumption data, stability data, maintainability data, and cost data of each of the first voiceprint recognition models based on a target source;

[0177] A first generating submodule, for generating, based on a preset scoring strategy, a performance score, a resource consumption score, a stability score, a maintainability score and a cost score of each of the first voiceprint recognition models corresponding to the performance data, resource consumption data, stability data, maintainability data and cost data of each of the first voiceprint recognition models;

[0178] A second generating submodule, configured to generate a model score of each of the first voiceprint recognition models based on the performance score, resource consumption score, stability score, maintainability score and cost score of each of the first voiceprint recognition models;

[0179] An acquisition submodule, used for acquiring the target model score with the highest value from all the model scores;

[0180] The first determination submodule is used to screen out a second voiceprint recognition model corresponding to the target model score from all the first voiceprint recognition models, and use the second voiceprint recognition model as the target voiceprint recognition model.

[0181] In some optional implementations of this embodiment, the second generation submodule includes:

[0182] An acquisition unit, used to acquire a preset weight setting algorithm;

[0183] A generating unit, configured to generate factor weights corresponding to the performance factor, the resource consumption factor, the stability factor, the maintainability factor and the cost factor respectively based on the weight setting algorithm;

[0184] A calling unit, used to call a preset score calculation formula;

[0185] A calculation unit is used to calculate the performance score, resource consumption score, stability score, maintainability score, cost score and factor weight of each of the first voiceprint recognition models based on the score calculation formula to obtain a model score of each of the first voiceprint recognition models.

[0186] In some optional implementations of this embodiment, the first acquisition module 305 includes:

[0187] The first calling submodule is used to call a preset preprocessing strategy library;

[0188] A matching submodule, configured to perform matching processing on the pre-processing strategy library based on the business scenario information, so as to query the pre-processing strategy library for a specified audio pre-processing strategy that matches the business scenario information;

[0189] The second determining submodule is configured to use the specified audio preprocessing strategy as a target audio preprocessing strategy.

[0190] In some optional implementations of this embodiment, the marking module 302 includes:

[0191] The second calling submodule is used to call a preset emotion recognition model;

[0192] The recognition submodule is used to input the telephone recording into the emotion recognition model, perform emotion recognition processing on the telephone recording through the emotion recognition model and output a corresponding emotion recognition result;

[0193] A labeling submodule, used for labeling the text with emotions based on the emotion recognition result to obtain the labeled text;

[0194] The third determination submodule is used to use the annotated text as the target text.

[0195] In some optional implementations of this embodiment, the voiceprint recognition device based on artificial intelligence further includes:

[0196] A second acquisition module is used to acquire a pre-collected telephone recording data set; wherein the telephone recording data set includes designated telephone recordings in various business scenarios and labels corresponding to the designated telephone recordings;

[0197] A preprocessing module, used to preprocess the telephone recording data set to obtain corresponding telephone recording sample data;

[0198] The second calling module is used to call the pre-trained language model;

[0199] A training module, used for training and fine-tuning the pre-trained language model based on the telephone recording sample data and a preset low-rank adapter to obtain a specified language model that meets expected conditions;

[0200] The first determination module is used to use the specified language model as the service identification model.

[0201] In some optional implementations of this embodiment, the voiceprint recognition device based on artificial intelligence further includes:

[0202] The third acquisition module is used to obtain a preset formatting strategy;

[0203] A formatting module, used to format the voiceprint recognition result based on the formatting strategy to obtain a corresponding target voiceprint recognition result;

[0204] A generating module, used to generate explanation information corresponding to the voiceprint recognition result;

[0205] The second determination module is used to determine the target display mode;

[0206] A sending module is used to send the target voiceprint recognition result and the explanation information to a preset downstream system based on the target display method.

[0207] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0208] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.

[0209] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0210] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk, smart memory card (Smart Med ia Card, SMC), secure digital (Secure Digital, SD) card, flash card (Flash Card), etc. equipped on the computer device 4. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the voiceprint recognition method based on artificial intelligence, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0211] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the artificial intelligence-based voiceprint recognition method.

[0212] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0213] The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned artificial intelligence-based voiceprint recognition method.

[0214] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0215] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A voiceprint recognition method based on artificial intelligence, characterized in that: The steps include: Obtaining a telephone recording to be processed, and transcribing the telephone recording to obtain a corresponding text; Performing emotion recognition on the telephone recording to obtain a corresponding emotion recognition result, and performing emotion annotation on the text based on the emotion recognition result to obtain a corresponding target text; Calling a pre-trained business identification model; wherein the business identification model is a model obtained by training a pre-trained language model based on pre-collected telephone recording sample data; Performing recognition processing on the target text based on the business recognition model to obtain business scenario information corresponding to the target text; Acquire a target audio preprocessing strategy corresponding to the business scenario information, and preprocess the telephone recording based on the target audio preprocessing strategy to obtain corresponding target voice data; Based on the preset screening strategy and the business scenario information, the preset multiple voiceprint recognition models are screened and processed to determine the corresponding target voiceprint recognition model; The target voiceprint recognition model is used to perform voiceprint recognition processing on the target speech data to obtain a corresponding voiceprint recognition result.

2. The voiceprint recognition method based on artificial intelligence according to claim 1 is characterized in that: The step of screening the preset multiple voiceprint recognition models based on the preset screening strategy and the business scenario information to determine the corresponding target voiceprint recognition model specifically includes: Acquire all the voiceprint recognition models, and select a first voiceprint recognition model that meets the business scenario information from all the voiceprint recognition models; Collecting performance data, resource consumption data, stability data, maintainability data, and cost data of each of the first voiceprint recognition models based on the target source; Generate, based on a preset scoring strategy, a performance score, a resource consumption score, a stability score, a maintainability score, and a cost score for each of the first voiceprint recognition models corresponding to the performance data, resource consumption data, stability data, maintainability data, and cost data of each of the first voiceprint recognition models; generating a model score for each of the first voiceprint recognition models based on the performance score, resource consumption score, stability score, maintainability score, and cost score of each of the first voiceprint recognition models; Obtain the target model score with the highest value from all the model scores; A second voiceprint recognition model corresponding to the target model score is screened out from all the first voiceprint recognition models, and the second voiceprint recognition model is used as the target voiceprint recognition model.

3. The voiceprint recognition method based on artificial intelligence according to claim 2 is characterized in that: The step of generating a model score of each of the first voiceprint recognition models based on the performance score, resource consumption score, stability score, maintainability score and cost score of each of the first voiceprint recognition models specifically includes: Get the preset weight setting algorithm; Generating factor weights corresponding to the performance factor, the resource consumption factor, the stability factor, the maintainability factor and the cost factor respectively based on the weight setting algorithm; Call the preset score calculation formula; Based on the score calculation formula, the performance score, resource consumption score, stability score, maintainability score, cost score and the factor weight of each of the first voiceprint recognition models are calculated and processed to obtain a model score of each of the first voiceprint recognition models.

4. The voiceprint recognition method based on artificial intelligence according to claim 1, characterized in that: The step of obtaining a target audio preprocessing strategy corresponding to the business scenario information specifically includes: Call the preset preprocessing strategy library; Performing matching processing on the preprocessing policy library based on the business scenario information, so as to query a specified audio preprocessing policy matching the business scenario information from the preprocessing policy library; The specified audio preprocessing strategy is used as the target audio preprocessing strategy.

5. The voiceprint recognition method based on artificial intelligence according to claim 1, characterized in that: The step of performing emotion recognition on the telephone recording to obtain a corresponding emotion recognition result, and performing emotion annotation on the text based on the emotion recognition result to obtain a corresponding target text specifically includes: Call the preset emotion recognition model; Inputting the telephone recording into the emotion recognition model, performing emotion recognition processing on the telephone recording through the emotion recognition model and outputting corresponding emotion recognition results; Performing emotion annotation on the text based on the emotion recognition result to obtain annotated text; The annotated text is used as the target text.

6. The voiceprint recognition method based on artificial intelligence according to claim 1, characterized in that: Before the step of calling the pre-trained service identification model, the method further includes: Acquire a pre-collected telephone recording data set; wherein the telephone recording data set includes designated telephone recordings in various business scenarios and labels corresponding to the designated telephone recordings; Preprocessing the telephone recording data set to obtain corresponding telephone recording sample data; Call the pre-trained language model; Based on the telephone recording sample data and the preset low-rank adapter, the pre-trained language model is trained and fine-tuned to obtain a specified language model that meets the expected conditions; The specified language model is used as the service identification model.

7. The voiceprint recognition method based on artificial intelligence according to claim 1, characterized in that: After the step of performing voiceprint recognition processing on the target voice data based on the target voiceprint recognition model to obtain a corresponding voiceprint recognition result, the method further includes: Get the preset formatting strategy; Formatting the voiceprint recognition result based on the formatting strategy to obtain a corresponding target voiceprint recognition result; Generating explanation information corresponding to the voiceprint recognition result; Determine the target presentation method; Based on the target display method, the target voiceprint recognition result and the interpretation information are sent to a preset downstream system.

8. A voiceprint recognition device based on artificial intelligence, characterized in that: include: A transcription module, used to obtain the telephone recording to be processed, and transcribe the telephone recording to obtain the corresponding text; A tagging module, used for performing emotion recognition on the telephone recording to obtain a corresponding emotion recognition result, and performing emotion tagging on the text based on the emotion recognition result to obtain a corresponding target text; A first calling module is used to call a pre-trained service identification model; wherein the service identification model is a model obtained by training a pre-trained language model based on pre-collected telephone recording sample data; A first recognition module, configured to perform recognition processing on the target text based on the business recognition model to obtain business scenario information corresponding to the target text; A first acquisition module is used to acquire a target audio preprocessing strategy corresponding to the business scenario information, and preprocess the telephone recording based on the target audio preprocessing strategy to obtain corresponding target voice data; A screening module, used to screen the preset multiple voiceprint recognition models based on the preset screening strategy and the business scenario information, and determine the corresponding target voiceprint recognition model; The second recognition module is used to perform voiceprint recognition processing on the target speech data based on the target voiceprint recognition model to obtain a corresponding voiceprint recognition result.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the voiceprint recognition method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps of the voiceprint recognition method based on artificial intelligence as described in any one of claims 1 to 7 are implemented.

Images