Voiceprint recognition method and device, computer equipment and storage medium
By using voiceprint recognition methods in the customer service system, combining historical and real-time voiceprint feature information, and using a similarity evaluation mechanism, the problem of low sound recognition efficiency and accuracy is solved, and higher recognition accuracy and reliability are achieved.
Patent Information
- Application Number
- CN202410834219.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-26
- Publication Date
- 2024-11-22
AI Technical Summary
In the prior art, when customer service personnel pick up customers in the line, due to the similar voices of different users, the sound recognition efficiency and accuracy are low, manual recording delays and omissions are large, and the cost is high.
By obtaining audio data based on the historical database for preprocessing, standard voiceprint feature information is extracted, and combined with real-time voiceprint feature information, a similarity evaluation mechanism is used, including similarity algorithm and voting and election method, to calculate similarity scores and generate voiceprint recognition Reports are sent to the user terminal.
It improves the efficiency and accuracy of sound recognition, reduces the false matching rate, reduces the possibility of misidentification, improves the reliability of recognition results, and is suitable for fast response application scenarios.
Smart Images

Figure CN119007729A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of speech recognition technology, and in particular to a voiceprint recognition method, device, computer equipment and storage medium. Background Art
[0002] When customer service staff were answering calls from customers, they found that the voices of different users were extremely similar and their speech was highly professional. Manual recording of people's voices and speech had extremely high delays and omissions, and the labor cost was extremely high.
[0003] Therefore, there is an urgent need to propose a voiceprint recognition method, device, computer equipment and storage medium that can improve the efficiency and accuracy of sound recognition. Summary of the invention
[0004] Based on this, it is necessary to provide a voiceprint recognition method, device, computer equipment and storage medium that can improve the efficiency and accuracy of sound recognition in response to the above technical problems.
[0005] In one aspect, a voiceprint recognition method is provided, the method comprising:
[0006] Based on the historical database, obtaining at least one audio data, and preprocessing the audio data to obtain standard voiceprint feature information;
[0007] Obtain real-time voiceprint feature information;
[0008] Determining a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on a similarity evaluation mechanism, wherein the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm, and comparing the similarity coefficients based on a voting method;
[0009] In response to detecting that the similarity score is greater than a preset threshold, a voiceprint recognition report is generated and sent to the user terminal.
[0010] Optionally, based on the historical database, at least one audio data is obtained, and the audio data is preprocessed to obtain standard voiceprint feature information including:
[0011] Acquire target audio data from a historical database, and extract multiple audio segments from the target audio data;
[0012] Performing noise reduction and vocal enhancement processing on the multiple audio clips, and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip;
[0013] The valid audio segment is input into a pre-built voiceprint recognition model to obtain the standard voiceprint feature information.
[0014] Optionally, after obtaining the standard voiceprint feature information, the method further includes:
[0015] Obtaining basic user information corresponding to the standard voiceprint feature information, and verifying the basic user information;
[0016] Determine the attribute label of the standard voiceprint feature information according to the verification result, and generate a unique identifier based on the attribute label;
[0017] Based on the unique identifier, the standard voiceprint feature information is stored in a corresponding audio database, and the audio database at least includes: a voiceprint archive library, a voiceprint group library and a customer complaint list library.
[0018] Optionally, a method for acquiring real-time voiceprint feature information includes:
[0019] Acquire user real-time audio data, and extract multiple audio clips from the user real-time audio data;
[0020] Performing noise reduction and vocal enhancement processing on the multiple audio clips, and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip;
[0021] The valid audio segment is input into a pre-built voiceprint recognition model to obtain the real-time voiceprint feature information.
[0022] Optionally, before determining the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the similarity evaluation mechanism, the solution further includes:
[0023] Determining the priority of an audio database for obtaining target standard voiceprint feature information according to the business scenario to which the real-time voiceprint feature information belongs;
[0024] According to the audio database priority and voiceprint comparison mode, the target standard voiceprint feature information is obtained in sequence to calculate the similarity score.
[0025] Optionally, before determining the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the similarity evaluation mechanism, the solution further includes:
[0026] Based on the audio database to which the target standard voiceprint feature information belongs, selecting a data analysis pipeline for determining a similarity score, and generating a data analysis task;
[0027] Distribute multiple data analysis tasks to corresponding data analysis pipelines, and calculate and determine a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the data analysis pipeline.
[0028] Optionally, based on a similarity evaluation mechanism, determining a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information includes:
[0029] Using a similarity algorithm, calculate the similarity between the feature vectors corresponding to the real-time voiceprint feature information and the target standard voiceprint feature information to obtain a target two-dimensional array, wherein the target two-dimensional array at least includes a similarity matrix and a similarity coefficient in the similarity matrix;
[0030] Determining a voting threshold based on the voting election law;
[0031] Based on the voting threshold, voting is performed on similarity coefficients in the similarity matrix that are greater than or equal to the voting threshold to obtain a number of votes;
[0032] The ratio of the number of votes to the number of similarity coefficients in the similarity matrix is calculated, and the ratio is defined as the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information. The number of similarity coefficients is obtained by the number of groups of voiceprint feature information.
[0033] In another aspect, a voiceprint recognition device is provided, the device comprising:
[0034] A standard voiceprint feature information acquisition module, used to acquire at least one audio data based on a historical database, and pre-process the audio data to obtain standard voiceprint feature information;
[0035] A real-time voiceprint feature information acquisition module is used to acquire real-time voiceprint feature information;
[0036] A similarity evaluation module, configured to determine a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on a similarity evaluation mechanism, wherein the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm and comparing the similarity coefficients based on a voting method;
[0037] The report generation module is used to generate a voiceprint recognition report and send it to the user terminal when it is detected that the similarity score is greater than a preset threshold.
[0038] In another aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the following steps are implemented:
[0039] Based on the historical database, obtaining at least one audio data, and preprocessing the audio data to obtain standard voiceprint feature information;
[0040] Obtain real-time voiceprint feature information;
[0041] Determining a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on a similarity evaluation mechanism, wherein the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm, and comparing the similarity coefficients based on a voting method;
[0042] In response to detecting that the similarity score is greater than a preset threshold, a voiceprint recognition report is generated and sent to the user terminal.
[0043] In another aspect, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
[0044] Based on the historical database, obtaining at least one audio data, and preprocessing the audio data to obtain standard voiceprint feature information;
[0045] Obtain real-time voiceprint feature information;
[0046] Determining a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on a similarity evaluation mechanism, wherein the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm, and comparing the similarity coefficients based on a voting method;
[0047] In response to detecting that the similarity score is greater than a preset threshold, a voiceprint recognition report is generated and sent to the user terminal.
[0048] The above-mentioned voiceprint recognition method, device, computer equipment and storage medium, the method includes: based on a historical database, obtaining at least one audio data, and preprocessing the audio data to obtain standard voiceprint feature information; obtaining real-time voiceprint feature information; based on a similarity evaluation mechanism, determining the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information, the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm, and comparing the similarity coefficients based on a voting method; in response to detecting that the similarity score is greater than a preset threshold, generating a voiceprint recognition report and sending it to the user terminal. The present application is based on a high-accuracy voiceprint feature recognition method, which can realize real-time and accurate extraction and analysis of sound features, can be applied to application scenarios that require rapid response, greatly reduces the false matching rate, effectively reduces the possibility of misidentification, and improves the reliability of the recognition results. BRIEF DESCRIPTION OF THE DRAWINGS
[0049] Figure 1 A diagram showing an application environment of a voiceprint recognition method in an embodiment;
[0050] Figure 2 A schematic diagram of the overall process of a voiceprint recognition method in an embodiment;
[0051] Figure 3 A schematic diagram of an audio preprocessing process of a voiceprint recognition method in one embodiment;
[0052] Figure 4 A schematic diagram of a voiceprint creation process of a voiceprint recognition method in one embodiment;
[0053] Figure 5 Another overall flow chart of a voiceprint recognition method in one embodiment;
[0054] Figure 6 A flowchart of data analysis in different modes of a voiceprint recognition method in one embodiment;
[0055] Figure 7 A flowchart of data analysis in different modes of a voiceprint recognition method in one embodiment;
[0056] Figure 8 A schematic diagram of voiceprint comparison in a voiceprint recognition method in one embodiment;
[0057] Fig. 9 A schematic diagram of similarity calculation results of a voiceprint recognition method in one embodiment;
[0058] Fig.10 A schematic diagram of the system structure of a voiceprint recognition method in an embodiment;
[0059] Fig.11 A schematic diagram of a similarity calculation process of a voiceprint recognition method in an embodiment;
[0060] Fig.12 is another flow chart of a voiceprint recognition method in one embodiment;
[0061] Fig.13 Creating a schematic diagram for a voiceprint comparison task of a voiceprint recognition method in one embodiment;
[0062] Fig.14 A schematic diagram of an identification report of a voiceprint identification method in an embodiment;
[0063] Fig.15 A schematic diagram of a recognition result of a voiceprint recognition method in an embodiment;
[0064] Fig.16 is a structural block diagram of a voiceprint recognition device in an embodiment;
[0065] Fig.17 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION
[0066] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0067] It should be understood that in the description of the present application, unless the context clearly requires otherwise, words such as "include", "comprises", and the like throughout the specification should be interpreted as inclusive rather than exclusive or exhaustive; that is, the meaning of "including but not limited to".
[0068] It should also be understood that the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, "plurality" means two or more.
[0069] It should be noted that the terms "S1", "S2", etc. are only used for the purpose of describing the steps, and do not specifically refer to the order or sequence, nor are they used to limit the present application. They are only for the convenience of describing the method of the present application, and cannot be understood as indicating the order of the steps. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in this field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present application.
[0070] In the personal loan business, users (borrowers) ask third-party agents to handle the platform's collection activities on their behalf, including malicious complaints and phishing complaints to the State Administration of Financial Supervision and Administration. The emergence of agent gangs threatens the platform's regulatory risks, damages the platform's economic interests, and damages personal economic interests.
[0071] Therefore, there is an urgent need to propose a voiceprint recognition method, device, computer equipment and storage medium that can improve the efficiency and accuracy of sound recognition.
[0072] According to the background technology, when customer service personnel answer the calls from customers, they find that the voices of different users are extremely similar and their speech is highly professional. The existing technology usually records people's voices and speech manually. This method has extremely high delays and omissions, and the labor cost is extremely high.
[0073] To solve the above technical problems, this application provides a voiceprint recognition method, device, equipment and storage medium. The voice similarity is used to determine that the person answering the phone for collection and comfort is not the real person, and it is inferred that there may be an agent involved. The suspected agent voice is compared with the historical data to try to dig out the agent gang, which provides an important guarantee for financial security.
[0074] The voiceprint recognition method provided in this application can be applied to Figure 1 In the application environment shown, the terminal 102 communicates with the data processing platform set on the server 104 through the network, wherein the terminal 102 can be but not limited to various personal computers, laptops, smart phones, tablet computers and portable wearable devices, and the server 104 can be implemented by an independent server or a server cluster composed of multiple servers.
[0075] In one embodiment, Figure 2 As shown, a voiceprint recognition method is provided, which is applied to Figure 1 The terminal in is used as an example to illustrate, including the following steps:
[0076] S1: Based on a historical database, obtain at least one audio data, and preprocess the audio data to obtain standard voiceprint feature information.
[0077] It should be noted that the historical database refers to the database that stores the user's original audio data and basic information. The preprocessing methods include denoising and splicing of audio data. The standard voiceprint feature information refers to the basic data used for subsequent comparison.
[0078] In some specific embodiments, Figure 3 and Figure 5 As shown, based on the historical database, at least one audio data is obtained, and the audio data is preprocessed to obtain standard voiceprint feature information including:
[0079] Obtain target audio data from a historical database, and extract multiple audio segments from the target audio data, wherein the target audio data refers to audio data of a certain user, such as call recordings, live conversations, etc., and multiple high-quality and effective audio sound segments can be intercepted from the target audio data through ASR (Automatic Speech Recognition), AIGC (Artificial Intelligence Generated Content), and other related technologies, wherein the sound segment extraction method is a common method and will not be repeated here;
[0080] Performing channel separation, noise reduction and vocal enhancement processing on the multiple audio clips (or audio fragments), and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip, wherein the noise reduction, vocal enhancement processing, splicing and transcoding of audio data are common means, which will not be repeated here, and the valid audio clips may include one or more, preferably 4 or more, so as to facilitate the subsequent calculation of similarity scores;
[0081] The valid audio clip is input into a pre-constructed voiceprint recognition model to obtain the standard voiceprint feature information, wherein the voiceprint recognition model can be constructed by a TDNN time-delay neural network or other deep neural networks that can be used for speech recognition, and a training set and a test set are constructed by a large amount of audio data to train and test the constructed voiceprint recognition model. When the test accuracy reaches the preset standard, the audio clip can be recognized based on the trained voiceprint recognition model. Each valid audio clip can obtain n groups of data containing 192 features. Finally, n*192 data are used as the voiceprint feature information of an audio data.
[0082] Specifically, in the process of customer complaints or other business processing, the user's environment, speech, emotions, etc. all affect the voiceprint recognition effect. In order to deal with the problem of inconsistent call audio quality caused by factors such as environment, recording equipment and user emotional fluctuations, this application takes a series of optimization measures to process the original audio, including: first, intercepting high-quality and effective sound segment audio in the user's audio data, intercepting the sound quality environment characteristics of black market agent complaints, performing speech recognition based on ASR and other technologies, intercepting high-quality audio segments, and performing noise reduction processing on the intercepted high-quality audio segments to remove white noise. At the same time, the audio data is processed with human voice enhancement and the audio is spliced. Subsequently, the audio is transcoded to obtain effective high-quality audio clips for subsequent processing; through the above-mentioned optimization measures, audio data such as call audio can be better processed, and high-quality audio clips can be extracted, thereby reducing the impact of factors such as environment and equipment on audio quality, thereby improving the accuracy and reliability of subsequent voiceprint recognition and other operations.
[0083] Furthermore, by using a voiceprint recognition model trained with a large amount of audio data, valid audio fragments are identified to obtain standard voiceprint feature information, that is, each audio can obtain n groups of data containing 192 features. Finally, n*192 data are used as the voiceprint features of an audio.
[0084] In some specific embodiments, Figure 4 As shown, after obtaining the standard voiceprint feature information, the method further includes:
[0085] Obtaining basic user information corresponding to the standard voiceprint feature information, and verifying the basic user information, wherein the basic user information may include relevant information such as mobile phone number, ID card and name, and the verification based on the basic user information is to confirm whether the user corresponding to the extracted standard voiceprint feature information is the policy user;
[0086] According to the verification result, the attribute label of the standard voiceprint feature information is determined, and based on the attribute label, a unique identifier, i.e., a voiceprint ID, is generated. If it is a policy user, a corresponding attribute label is attached to the user. If the attribute label is a policy user, the generated unique identifier may be a policy user number, etc. If it is not a policy user, a corresponding attribute label is attached to the user. If the attribute label is a non-policy user, the generated unique identifier may be a non-policy user sorting number, etc. In addition, if the user is a customer complaint user, an identifier for identifying the customer complaint user, such as ks, etc., is added to the above unique identifier, so as to subsequently allocate the standard voiceprint feature information of multiple users to the corresponding audio database for storage;
[0087] Based on the unique identifier, the standard voiceprint feature information is stored in a corresponding audio database, and the audio database at least includes: a voiceprint archive, a voiceprint group library and a customer complaint list library, wherein the voiceprint archive stores all audio data of customers of the customer service system, such as user A audio, user B audio, etc., the customer complaint list stores the audio data of all customer complaint users, and the voiceprint group library stores historical audio data with fraud risks identified by voiceprints. The data in the above databases can be stored in corresponding kafka (open source stream processing platform) pipelines, i.e., data processing pipelines. For example, the customer complaint list library is stored in a platform dedicated to processing customer complaint data. Based on a dedicated platform, it is used for subsequent comparison of real-time customer complaint voice data to improve data processing efficiency. Users can select a suitable audio library for voiceprint comparison according to actual business scenarios.
[0088] Specifically, Figure 4 As shown, the process of creating a voiceprint includes verifying the user's basic information in the Diting anti-fraud system to generate a voiceprint ID, extracting the user's basic audio information, such as audio recording time, etc., uploading the extracted audio data to the object storage service (OSS), that is, storing it in the audio database, and further submitting a voiceprint asynchronous creation task to asynchronously call relevant data processing software to generate a voiceprint, that is, standard voiceprint feature information, and compressing and encrypting the generated standard voiceprint feature information and storing it in the corresponding audio database. The original audio file and related high-quality audio clips will be stored in the object storage service (OSS) as the basic data for subsequent voiceprint comparison.
[0089] S2: Obtain real-time voiceprint feature information.
[0090] It should be noted that the real-time voiceprint feature information is the voiceprint feature information obtained after processing the user's instant call audio, and the processing method is the same as the preprocessing method in step S1.
[0091] In some specific implementations, the method for acquiring real-time voiceprint feature information includes:
[0092] Acquire user real-time audio data, extract multiple audio clips from the user real-time audio data, where Fig.10 As shown, the user's real-time audio data may include collection, complaint or other related voice data. This step also uses ASR and other technologies to perform speech recognition and capture multiple high-quality audio segments.
[0093] The multiple audio clips are subjected to noise reduction and vocal enhancement processing, and the processed multiple audio clips are spliced and transcoded to obtain at least one valid audio clip, wherein similarly, noise reduction, vocal enhancement processing, splicing and transcoding of audio data are common means, which will not be repeated here, and the valid audio clips may include one or more, preferably 4 or more, such as Fig.12 As shown, it includes voiceprint frequency bands 1 to 4, so as to facilitate the subsequent calculation of similarity scores;
[0094] The valid audio clip is input into a pre-constructed voiceprint recognition model to obtain the real-time voiceprint feature information, wherein, similarly, the voiceprint recognition model can be constructed by a TDNN time-delay neural network or other deep neural networks that can be used for speech recognition, and a training set and a test set are constructed by a large amount of audio data to train and test the constructed voiceprint recognition model. When the test accuracy reaches the preset standard, the audio clip can be recognized based on the trained voiceprint recognition model. Each valid audio clip can obtain n groups of data containing 192 features. Finally, the n*192 data is used as the voiceprint feature information of an audio data. Exemplarily, if n is 4 as mentioned above, the voiceprint feature information is 4*192 data.
[0095] S3: Based on a similarity evaluation mechanism, determine a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information.
[0096] It should be noted that the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm, and comparing the similarity coefficients based on a voting method, wherein the similarity algorithm may be a cosine similarity algorithm or other algorithms for calculating similarity, such as Euclidean distance, text matching model, etc. This embodiment preferably uses a cosine similarity algorithm to calculate the similarity between the real-time voiceprint feature information and the target standard voiceprint feature information.
[0097] In some specific implementations, before determining the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the similarity evaluation mechanism, the method further includes:
[0098] like Fig.11 As shown, according to the business scenario to which the real-time voiceprint feature information belongs, the priority of the audio database used to obtain the target standard voiceprint feature information is determined, wherein the business scenario may include customer complaint scenarios, ordinary business processing scenarios, and risk control scenarios, and different scenarios may have different audio database data for comparison. For example, in the customer complaint scenario, the voiceprint feature information in the customer complaint list library is generally used for comparison first. If the comparison in the customer complaint list library is not successful, other data in the voiceprint archive library is used for comparison. In the risk control scenario, the voiceprint feature information in the voiceprint group library is used for comparison first. If the comparison is not successful, other data in the voiceprint archive library is used for comparison, and so on. Therefore, the audio database corresponding to the business scenario has a higher priority, and the voiceprint archive library has a lower priority.
[0099] According to the audio database priority and voiceprint comparison mode, the target standard voiceprint feature information is obtained in sequence to calculate the similarity score, where: Fig.13 As shown, the voiceprint comparison mode includes 5 modes, including standard mode, custom threshold mode, any match mode, full mode and personal comparison mode. Among them, in standard mode, the system compares according to the mode threshold set by the comparison system to identify the target audio and similar audio in the audio database. At the same time, the system also supports user-defined threshold mode so that users can adjust the threshold at any time according to the actual usage scenario. The any match mode is designed to quickly search the audio database to determine whether there is audio similar to the target audio. The full mode provides a 360-degree comprehensive search without dead ends to find all audio similar to the target audio in the audio library. The personal comparison mode can quickly determine whether two calls are from the same person and whether there is any fraudulent agency behavior. These modes can be selected by the user or directly selected by the system according to the actual scenario to meet different data identification and comparison needs.
[0100] In some specific implementations, before determining the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the similarity evaluation mechanism, the method further includes:
[0101] Based on the audio database to which the target standard voiceprint feature information belongs, a data analysis pipeline for determining a similarity score is selected, and a data analysis task is generated, wherein the data analysis pipeline is the above-mentioned kafka pipeline, and each pipeline only performs comparative analysis on data of the same type. For example, if the pipeline is used to analyze data in a customer complaint scenario, it only processes the comparison operation between the customer complaint audio data and the data in the customer complaint list library;
[0102] Distribute multiple data analysis tasks to corresponding data analysis pipelines, and calculate and determine a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the data analysis pipeline.
[0103] Specifically, in the process of massive voiceprint comparison, this embodiment divides and segments the voiceprint features into sharding segments, slices the same type of data into the same process pipeline based on multi-dimensional routing data, realizes efficient audio task distribution, sends the sliced data to the Kafka pipeline, and quickly identifies countless sliced data based on the Kafka pipeline. Through this sharded management and free control of the shard size, it is ensured that the data is accurately divided and sent to the appropriate data analysis pipeline, and the optimal solution for the comparison speed is achieved for different scenarios. It can realize efficient parallel processing of massive voiceprint feature information, thereby improving the comparison speed and recognition accuracy.
[0104] Further, such as Figure 6-7 As shown in the figure, the processing flow of different Kafka pipelines for different voiceprint comparison modes includes:
[0105] (1) If the voiceprint comparison mode is the standard mode, the custom threshold mode, or the full mode, determine the data range of the current database, divide the ID interval, and distribute the data analysis tasks. For example, the divided data range is 1-100...900-988, etc. The comparison and analysis process of the kafka pipeline is: compare the valid voiceprint feature information, accumulate the number of IDs in this interval num and save it to redis (log or database), and when num is greater than or equal to maxlds, update the task completion status to "Comparison completed";
[0106] (2) If the voiceprint comparison mode is any mode, divide the data range, divide the ID interval and distribute the data analysis tasks. For example, the divided data range can be 1-100...900-988, etc. The comparison and analysis process of the kafka pipeline is: compare the valid voiceprint feature information. When the similarity is greater than the preset threshold, update the flag=Y in redis. Otherwise, save the number of IDs num in this interval to redis (log or database). When num is greater than or equal to maxlds, update the task completion status to "comparison completed".
[0107] In some specific embodiments, Figure 5 As shown, based on the similarity evaluation mechanism, determining the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information includes:
[0108] The similarity algorithm is used to calculate the similarity between the feature vectors corresponding to the real-time voiceprint feature information and the target standard voiceprint feature information to obtain a target two-dimensional array. The target two-dimensional array at least includes a similarity matrix and a similarity coefficient in the similarity matrix. The preferred similarity algorithm in this embodiment is a cosine similarity algorithm. The cosine similarity algorithm is a method commonly used to measure the similarity between vectors. The cosine value of the angle between the vectors is calculated to determine their similarity. In the voiceprint feature recognition process, the cosine similarity algorithm is used to evaluate the similarity between feature vectors. For example, if the vector matrix of the two voiceprint feature information is (n, 192), if n is 4, the calculated similarity matrix and the similarity coefficient in the similarity matrix are as follows: Fig. 9 As shown;
[0109] Based on the voting election method, determine a voting threshold, wherein the voting threshold can be set according to actual needs, such as 0.6;
[0110] Based on the voting threshold, similarity coefficients in the similarity matrix that are greater than or equal to the voting threshold are voted to obtain the number of votes. For example, as shown in the above similarity matrix and its similarity coefficient and the voting threshold of 0.6, it can be seen that the number of similarity coefficients in the similarity matrix that are greater than or equal to the voting threshold is 12, that is, the number of votes is 12;
[0111] Calculate the ratio of the number of votes to the number of similarity coefficients in the similarity matrix, define the ratio as the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information, the number of similarity coefficients is obtained by the number of groups of voiceprint feature information, such as when the real-time voiceprint feature information and the target standard voiceprint feature information are both 4 groups, the number of similarity coefficients is 16, then the ratio between the two is 0.75, and 0.75 is the similarity score.
[0112] For example, Figure 8 As shown, Shuke extracts voiceprints based on the original audio 1 and the comparison audio 2 respectively. If n=4, there is a section of audio with poor quality, so there are some low similarities, with an average value of 0.59. If the average value is compared with the set threshold of 0.6, and it is less than 0.6, then this group of comparison results are judged not to be the same person. For the audio voiceprint comparison of customer complaints, if it is calculated based on the voting election method, the final result is 0.75, then this group of comparison results are judged to be the same person, which is closer to the actual situation, and it is possible to directly find audio segments with high similarity from the comparison analysis results, which is convenient for manual verification and processing.
[0113] Furthermore, the process of determining the similarity score based on the similarity evaluation mechanism includes: extracting a 4*192 feature data from each audio segment of the original audio 1 and the comparison audio 2, and the two audio segments are two 4*192 feature data, using the cosine similarity algorithm to evaluate the similarity between the feature vectors, and calculating the similarity of the two 4*192 feature data, the following is obtained: Fig. 9 The data shown is stored in a two-dimensional array in the database. According to the set voting threshold (such as 0.6), the coefficients in the similarity matrix are compared with the voting threshold one by one. If the number is greater than or equal to the voting threshold, a vote is cast. Fig. 9 The number shown in is 12 votes, and the ratio between it and the total similarity coefficient is 12 / 16=0.75. 0.75 is the final similarity of the voiceprint feature information of the original audio 1 and the comparison audio 2 calculated.
[0114] In the above implementation, based on the cosine similarity algorithm, the similarity between multi-dimensional feature vectors can be effectively measured. Combined with the voting method, the similarities of different dimensions can be comprehensively considered, avoiding the limitations of a single feature dimension; when calculating the similarity, the cosine similarity algorithm has a certain anti-interference ability for some noise data, and can better handle noise and changes in sound signals; combined with the voting method, the similarity information of multiple voiceprint feature vectors can be used to better consider the contribution of different features to the voiceprint similarity, thereby improving the accuracy of recognition; combining the cosine similarity algorithm and the voting method can enhance the robustness of the voiceprint similarity calculation, reduce the sensitivity to outliers and noise, and improve the robustness of the voiceprint recognition system.
[0115] S4: In response to detecting that the similarity score is greater than a preset threshold, a voiceprint recognition report is generated and sent to the user terminal.
[0116] It should be noted that the preset threshold can be set according to actual needs, such as 0.7. When the similarity score is greater than 0.7, a corresponding voiceprint recognition report is generated and sent to the user terminal for the user to manually review again to correct the result, or to issue a fraud warning through the voiceprint recognition report to help customer service personnel identify and respond to fraudulent behavior, and provide strong support for customer complaint handling and anti-fraud work. Among them, the voiceprint recognition report is Fig.14 As shown, it includes the comparison mode, comparison library, creation time, comparison details and other related information. In addition, Fig.15 As shown, when the system determines that two videos are extremely similar, but the manual determination is extremely different, further verification and confirmation are required to improve the accuracy of recognition.
[0117] In the above-mentioned voiceprint recognition method, the method includes: based on a historical database, obtaining at least one audio data, and preprocessing the audio data to obtain standard voiceprint feature information; obtaining real-time voiceprint feature information; based on a similarity evaluation mechanism, determining the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information, the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm, and comparing the similarity coefficient based on a voting election method; in response to detecting that the similarity score is greater than a preset threshold, generating a voiceprint recognition report and sending it to the user terminal. The present application is based on a voiceprint model trained with a large amount of data, combined with a variety of audio processing methods and an audio data partition pipeline sending management mode, as well as 5 audio comparison modes and 3 audio databases, to quickly and intelligently identify user voices, and use a cosine similarity algorithm and voting elections to accurately calculate the similarity of voiceprint features, providing a more intelligent and convenient risk identification method for fraud security scenarios, improving the recognition accuracy and efficiency in the process of customer service black industry complaint identification, and realizing efficient fraud gang detection on the basis of low business personnel investment, reducing the property losses of black industry customer complaints to institutions and improving the work efficiency of customer service personnel.
[0118] This application is based on a highly accurate voiceprint feature recognition method, which can realize real-time and accurate extraction and analysis of sound features. It can be applied to application scenarios that require rapid response, greatly reduces the false match rate, effectively reduces the possibility of misidentification, and improves the reliability of recognition results.
[0119] It should be understood that although Figure 2-Figure 7 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2-Figure 7 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.
[0120] In one embodiment, Fig.16 As shown, a voiceprint recognition device is provided, including: a standard voiceprint feature information acquisition module, a real-time voiceprint feature information acquisition module, a similarity evaluation module and a report generation module, wherein:
[0121] A standard voiceprint feature information acquisition module, used to acquire at least one audio data based on a historical database, and pre-process the audio data to obtain standard voiceprint feature information;
[0122] A real-time voiceprint feature information acquisition module is used to acquire real-time voiceprint feature information;
[0123] A similarity evaluation module, configured to determine a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on a similarity evaluation mechanism, wherein the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm and comparing the similarity coefficients based on a voting method;
[0124] The report generation module is used to generate a voiceprint recognition report and send it to the user terminal when it is detected that the similarity score is greater than a preset threshold.
[0125] As a preferred implementation, in the embodiment of the present invention, the standard voiceprint feature information acquisition module is specifically used for:
[0126] Acquire target audio data from a historical database, and extract multiple audio segments from the target audio data;
[0127] Performing noise reduction and vocal enhancement processing on the multiple audio clips, and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip;
[0128] The valid audio segment is input into a pre-built voiceprint recognition model to obtain the standard voiceprint feature information.
[0129] As a preferred implementation, in the embodiment of the present invention, the device further includes a data storage module, and the data storage module is specifically used for:
[0130] Obtaining basic user information corresponding to the standard voiceprint feature information, and verifying the basic user information;
[0131] Determine the attribute label of the standard voiceprint feature information according to the verification result, and generate a unique identifier based on the attribute label;
[0132] Based on the unique identifier, the standard voiceprint feature information is stored in a corresponding audio database, and the audio database at least includes: a voiceprint archive library, a voiceprint group library and a customer complaint list library.
[0133] As a preferred implementation, in the embodiment of the present invention, the real-time voiceprint feature information acquisition module is specifically used for:
[0134] Acquire user real-time audio data, and extract multiple audio clips from the user real-time audio data;
[0135] Performing noise reduction and vocal enhancement processing on the multiple audio clips, and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip;
[0136] The valid audio segment is input into a pre-built voiceprint recognition model to obtain the real-time voiceprint feature information.
[0137] As a preferred implementation, in the embodiment of the present invention, the device further includes a target standard voiceprint feature information acquisition module, and the target standard voiceprint feature information acquisition module is specifically used to:
[0138] Determining the priority of an audio database for obtaining target standard voiceprint feature information according to the business scenario to which the real-time voiceprint feature information belongs;
[0139] According to the audio database priority and voiceprint comparison mode, the target standard voiceprint feature information is obtained in sequence to calculate the similarity score.
[0140] As a preferred implementation, in the embodiment of the present invention, the device further includes a data analysis pipeline determination module, and the data analysis pipeline determination module is specifically used to:
[0141] Based on the audio database to which the target standard voiceprint feature information belongs, selecting a data analysis pipeline for determining a similarity score, and generating a data analysis task;
[0142] Distribute multiple data analysis tasks to corresponding data analysis pipelines, and calculate and determine a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the data analysis pipeline.
[0143] As a preferred implementation, in the embodiment of the present invention, the similarity evaluation module is specifically used for:
[0144] Using a similarity algorithm, calculate the similarity between the feature vectors corresponding to the real-time voiceprint feature information and the target standard voiceprint feature information to obtain a target two-dimensional array, wherein the target two-dimensional array at least includes a similarity matrix and a similarity coefficient in the similarity matrix;
[0145] Determining a voting threshold based on the voting election law;
[0146] Based on the voting threshold, voting is performed on similarity coefficients in the similarity matrix that are greater than or equal to the voting threshold to obtain a number of votes;
[0147] The ratio of the number of votes to the number of similarity coefficients in the similarity matrix is calculated, and the ratio is defined as the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information. The number of similarity coefficients is obtained by the number of groups of voiceprint feature information.
[0148] For the specific definition of the voiceprint recognition device, please refer to the definition of the voiceprint recognition method above, which will not be repeated here. Each module in the above voiceprint recognition device can be implemented in whole or in part by software, hardware and their combination. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0149] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.17 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a voiceprint recognition method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse, etc.
[0150] Those skilled in the art will understand that Fig.17 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0151] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:
[0152] S1: Based on a historical database, obtaining at least one audio data, and preprocessing the audio data to obtain standard voiceprint feature information;
[0153] S2: Obtain real-time voiceprint feature information;
[0154] S3: determining a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on a similarity evaluation mechanism, wherein the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm, and comparing the similarity coefficients based on a voting method;
[0155] S4: In response to detecting that the similarity score is greater than a preset threshold, a voiceprint recognition report is generated and sent to the user terminal.
[0156] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0157] Acquire target audio data from a historical database, and extract multiple audio segments from the target audio data;
[0158] Performing noise reduction and vocal enhancement processing on the multiple audio clips, and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip;
[0159] The valid audio segment is input into a pre-built voiceprint recognition model to obtain the standard voiceprint feature information.
[0160] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0161] Obtaining basic user information corresponding to the standard voiceprint feature information, and verifying the basic user information;
[0162] Determine the attribute label of the standard voiceprint feature information according to the verification result, and generate a unique identifier based on the attribute label;
[0163] Based on the unique identifier, the standard voiceprint feature information is stored in a corresponding audio database, and the audio database at least includes: a voiceprint archive library, a voiceprint group library and a customer complaint list library.
[0164] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0165] Acquire user real-time audio data, and extract multiple audio clips from the user real-time audio data;
[0166] Performing noise reduction and vocal enhancement processing on the multiple audio clips, and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip;
[0167] The valid audio segment is input into a pre-built voiceprint recognition model to obtain the real-time voiceprint feature information.
[0168] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0169] Determining the priority of an audio database for obtaining target standard voiceprint feature information according to the business scenario to which the real-time voiceprint feature information belongs;
[0170] According to the audio database priority and voiceprint comparison mode, the target standard voiceprint feature information is obtained in sequence to calculate the similarity score.
[0171] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0172] Based on the audio database to which the target standard voiceprint feature information belongs, selecting a data analysis pipeline for determining a similarity score, and generating a data analysis task;
[0173] Distribute multiple data analysis tasks to corresponding data analysis pipelines, and calculate and determine a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the data analysis pipeline.
[0174] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:
[0175] Using a similarity algorithm, calculate the similarity between the feature vectors corresponding to the real-time voiceprint feature information and the target standard voiceprint feature information to obtain a target two-dimensional array, wherein the target two-dimensional array at least includes a similarity matrix and a similarity coefficient in the similarity matrix;
[0176] Determining a voting threshold based on the voting election law;
[0177] Based on the voting threshold, voting is performed on similarity coefficients in the similarity matrix that are greater than or equal to the voting threshold to obtain a number of votes;
[0178] The ratio of the number of votes to the number of similarity coefficients in the similarity matrix is calculated, and the ratio is defined as the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information. The number of similarity coefficients is obtained by the number of groups of voiceprint feature information.
[0179] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:
[0180] S1: Based on a historical database, obtaining at least one audio data, and preprocessing the audio data to obtain standard voiceprint feature information;
[0181] S2: Obtain real-time voiceprint feature information;
[0182] S3: determining a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on a similarity evaluation mechanism, wherein the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm, and comparing the similarity coefficients based on a voting method;
[0183] S4: In response to detecting that the similarity score is greater than a preset threshold, a voiceprint recognition report is generated and sent to the user terminal.
[0184] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0185] Acquire target audio data from a historical database, and extract multiple audio segments from the target audio data;
[0186] Performing noise reduction and vocal enhancement processing on the multiple audio clips, and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip;
[0187] The valid audio segment is input into a pre-built voiceprint recognition model to obtain the standard voiceprint feature information.
[0188] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0189] Obtaining basic user information corresponding to the standard voiceprint feature information, and verifying the basic user information;
[0190] Determine the attribute label of the standard voiceprint feature information according to the verification result, and generate a unique identifier based on the attribute label;
[0191] Based on the unique identifier, the standard voiceprint feature information is stored in a corresponding audio database, and the audio database at least includes: a voiceprint archive library, a voiceprint group library and a customer complaint list library.
[0192] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0193] Acquire user real-time audio data, and extract multiple audio clips from the user real-time audio data;
[0194] Performing noise reduction and vocal enhancement processing on the multiple audio clips, and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip;
[0195] The valid audio segment is input into a pre-built voiceprint recognition model to obtain the real-time voiceprint feature information.
[0196] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0197] Determining the priority of an audio database for obtaining target standard voiceprint feature information according to the business scenario to which the real-time voiceprint feature information belongs;
[0198] According to the audio database priority and voiceprint comparison mode, the target standard voiceprint feature information is obtained in sequence to calculate the similarity score.
[0199] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0200] Based on the audio database to which the target standard voiceprint feature information belongs, selecting a data analysis pipeline for determining a similarity score, and generating a data analysis task;
[0201] Distribute multiple data analysis tasks to corresponding data analysis pipelines, and calculate and determine a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the data analysis pipeline.
[0202] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:
[0203] Using a similarity algorithm, calculate the similarity between the feature vectors corresponding to the real-time voiceprint feature information and the target standard voiceprint feature information to obtain a target two-dimensional array, wherein the target two-dimensional array at least includes a similarity matrix and a similarity coefficient in the similarity matrix;
[0204] Determining a voting threshold based on the voting election law;
[0205] Based on the voting threshold, voting is performed on similarity coefficients in the similarity matrix that are greater than or equal to the voting threshold to obtain a number of votes;
[0206] The ratio of the number of votes to the number of similarity coefficients in the similarity matrix is calculated, and the ratio is defined as the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information. The number of similarity coefficients is obtained by the number of groups of voiceprint feature information.
[0207] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0208] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0209] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application.
Claims
1. A voiceprint recognition method, characterized in that: The method comprises: Based on the historical database, obtaining at least one audio data, and preprocessing the audio data to obtain standard voiceprint feature information; Obtain real-time voiceprint feature information; Determining a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on a similarity evaluation mechanism, wherein the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm, and comparing the similarity coefficients based on a voting method; In response to detecting that the similarity score is greater than a preset threshold, a voiceprint recognition report is generated and sent to the user terminal.
2. The voiceprint recognition method according to claim 1, characterized in that: Based on the historical database, at least one audio data is obtained, and the audio data is preprocessed to obtain standard voiceprint feature information including: Acquire target audio data from a historical database, and extract multiple audio segments from the target audio data; Performing noise reduction and vocal enhancement processing on the multiple audio clips, and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip; The valid audio segment is input into a pre-built voiceprint recognition model to obtain the standard voiceprint feature information.
3. The voiceprint recognition method according to claim 1 or 2, characterized in that: After obtaining the standard voiceprint feature information, the method further includes: Obtaining basic user information corresponding to the standard voiceprint feature information, and verifying the basic user information; Determine the attribute label of the standard voiceprint feature information according to the verification result, and generate a unique identifier based on the attribute label; Based on the unique identifier, the standard voiceprint feature information is stored in a corresponding audio database, and the audio database at least includes: a voiceprint archive library, a voiceprint group library and a customer complaint list library.
4. The voiceprint recognition method according to claim 1, characterized in that: The method for obtaining real-time voiceprint feature information includes: Acquire user real-time audio data, and extract multiple audio clips from the user real-time audio data; Performing noise reduction and vocal enhancement processing on the multiple audio clips, and splicing and transcoding the processed multiple audio clips to obtain at least one valid audio clip; The valid audio segment is input into a pre-built voiceprint recognition model to obtain the real-time voiceprint feature information.
5. The voiceprint recognition method according to claim 1, characterized in that: Before determining the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the similarity evaluation mechanism, the method further includes: Determining the priority of an audio database for obtaining target standard voiceprint feature information according to the business scenario to which the real-time voiceprint feature information belongs; According to the audio database priority and voiceprint comparison mode, the target standard voiceprint feature information is obtained in sequence to calculate the similarity score.
6. The voiceprint recognition method according to claim 5, characterized in that: Before determining the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the similarity evaluation mechanism, the method further includes: Based on the audio database to which the target standard voiceprint feature information belongs, selecting a data analysis pipeline for determining a similarity score, and generating a data analysis task; Distribute multiple data analysis tasks to corresponding data analysis pipelines, and calculate and determine a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on the data analysis pipeline.
7. The voiceprint recognition method according to claim 1, characterized in that: Based on the similarity evaluation mechanism, determining the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information includes: Using a similarity algorithm, calculate the similarity between the feature vectors corresponding to the real-time voiceprint feature information and the target standard voiceprint feature information to obtain a target two-dimensional array, wherein the target two-dimensional array at least includes a similarity matrix and a similarity coefficient in the similarity matrix; Determining a voting threshold based on the voting election law; Based on the voting threshold, voting is performed on similarity coefficients in the similarity matrix that are greater than or equal to the voting threshold to obtain a number of votes; The ratio of the number of votes to the number of similarity coefficients in the similarity matrix is calculated, and the ratio is defined as the similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information. The number of similarity coefficients is obtained by the number of groups of voiceprint feature information.
8. A voiceprint recognition device, characterized in that: The device comprises: A standard voiceprint feature information acquisition module, used to acquire at least one audio data based on a historical database, and pre-process the audio data to obtain standard voiceprint feature information; A real-time voiceprint feature information acquisition module is used to acquire real-time voiceprint feature information; A similarity evaluation module, configured to determine a similarity score between the real-time voiceprint feature information and the target standard voiceprint feature information based on a similarity evaluation mechanism, wherein the similarity evaluation mechanism includes determining a similarity coefficient based on a similarity algorithm and comparing the similarity coefficients based on a voting method; The report generation module is used to generate a voiceprint recognition report and send it to the user terminal when it is detected that the similarity score is greater than a preset threshold.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.
Citation Information
Cited By
Audio and video acquisition equipment data management system based on artificial intelligence
CN119513355A
Digital wireless audio signal noise reduction optimization method based on Matlab
CN119694331A
Audio recognition method, device, equipment, medium and product
CN121768401A
