A voiceprint registration method, device, computer equipment and storage medium

Abstract

The present application relates to the field of audio processing technology, and discloses a voiceprint registration method, device, computer equipment and storage medium. The voiceprint registration method of the present application obtains an ambient sound detection signal to determine whether the current ambient sound meets the voiceprint registration conditions. When the current ambient sound meets the voiceprint registration conditions, the first audio segment is recorded to obtain the first voiceprint information and the first feature information. When the first feature information reaches the decibel preset condition, the second audio segment is recorded to obtain the second voiceprint information and the second feature information. When the second feature information reaches the decibel preset condition, the first voiceprint feature of the first audio segment and the second voiceprint feature of the second audio segment are extracted, and the matching degree is calculated, and the audio segment that meets the preset matching degree is stored in the voiceprint library as the user's voiceprint. The voiceprint registration method of the present application enables the audio segment stored in the voiceprint library to have a strong anti-interference ability, ensures the integrity, reliability and accuracy of the voiceprint registration, and finally obtains a high-quality voiceprint.

Description

Technical Field

[0001] The present application relates to the field of audio processing technology, and in particular to a voiceprint registration method, device, computer equipment and storage medium. Background Art

[0002] With the continuous development of voiceprint recognition technology, confirming individual identity by analyzing the speaker's acoustic characteristics (such as frequency, rhythm, intensity, etc.) has become an indispensable part of the voiceprint recognition field. The role of this technology is increasingly being applied to video conferencing scenarios such as remote work, online education, and telemedicine. However, due to the high complexity of the field of voiceprint recognition technology, current voiceprint recognition systems usually use the signal-to-noise ratio calculated from the ambient audio as a reference. However, it does not take into account that the environment may change when the user is recording; it also ignores the problem that sudden noise or clipping that may occur during audio recording affects the accuracy of voiceprint features.

[0003] Therefore, how to obtain high-quality voiceprint data in an open and dynamically changing environment, and how to improve the anti-interference ability of voiceprint features while ensuring the stability and consistency of voiceprint features, have become important issues that need to be solved urgently. Summary of the invention

[0004] In view of this, the embodiments of the present application provide a voiceprint registration method, apparatus, computer equipment and storage medium, which can effectively solve the problem of environmental changes affecting the extraction accuracy that users may encounter when recording, and the problem of sudden noise or clipping during audio recording affecting the accuracy of voiceprint features.

[0005] In a first aspect, an embodiment of the present application provides a voiceprint registration method, including:

[0006] Obtain the ambient sound detection signal to determine whether the current ambient sound meets the voiceprint registration conditions;

[0007] When the current ambient sound meets the voiceprint registration condition, recording a first audio segment;

[0008] processing the first audio segment to obtain first voiceprint information and first feature information from the first audio segment;

[0009] When the first characteristic information reaches a preset decibel condition, recording a second audio segment;

[0010] processing the second audio segment to obtain second voiceprint information and second feature information from the second audio segment;

[0011] When the second characteristic information reaches the decibel preset condition, the first voiceprint feature of the first audio segment and the second voiceprint feature of the second audio segment are extracted, and the matching degree between the first voiceprint feature and the second voiceprint feature is calculated, and the audio segment that meets the preset matching degree is stored in the voiceprint library as the user voiceprint.

[0012] In some embodiments, before processing the first audio segment to obtain the first voiceprint information and the first feature information from the first audio segment, the method includes:

[0013] Obtaining a registration instruction from a user, and detecting the sound pressure decibel value of the current ambient sound according to the registration instruction;

[0014] Determine whether the sound pressure decibel value of the current ambient sound meets the preset recording condition;

[0015] When the sound pressure decibel value of the current ambient sound reaches the preset recording condition, the first audio segment is recorded.

[0016] In some embodiments, the processing the first audio segment to obtain first voiceprint information and first feature information from the first audio segment includes:

[0017] Preprocessing the first audio segment, and inputting a first feature vector group obtained after the preprocessing into a neural network model for training to obtain first voiceprint information in the first audio segment, wherein the first voiceprint information includes first valid speech duration information and first invalid noise duration information;

[0018] When the first effective voice duration information meets the first duration preset condition, obtaining first characteristic information;

[0019] The obtaining of the first characteristic information includes: calculating the root mean square of the first valid speech duration information and the first invalid noise duration information to obtain the first characteristic information.

[0020] In some embodiments, the processing the second audio segment to obtain second voiceprint information and second feature information from the second audio segment includes:

[0021] When the first characteristic information meets the decibel preset condition, recording a second audio segment;

[0022] Preprocessing the second audio segment, and inputting a second feature vector group obtained after the preprocessing into a neural network model for training to obtain second voiceprint information in the second audio segment, wherein the second voiceprint information includes second valid speech duration information and second invalid noise duration information;

[0023] When the second effective voice duration information meets the second duration preset condition, obtaining second characteristic information;

[0024] The obtaining of the second characteristic information includes: calculating the root mean square of the second valid speech duration information and the second invalid noise duration information to obtain the second characteristic information.

[0025] In some embodiments, the extracting a first voiceprint feature of the first audio segment during processing of the first audio segment includes:

[0026] The first valid speech duration information in the first audio segment is detected to obtain the window sound pressure information of the first audio segment; the window sound pressure information is compared with the first invalid noise duration information to obtain the first voiceprint feature of the first audio segment.

[0027] In some embodiments, extracting the second voiceprint feature of the second audio segment includes:

[0028] The effective duration information in the second audio segment is detected to obtain the window sound pressure information of the second audio segment; the window sound pressure information of the second audio segment is compared with the second invalid noise duration information to obtain the second voiceprint feature of the second audio segment.

[0029] In a second aspect, an embodiment of the present application provides a voiceprint registration device, including:

[0030] The voiceprint registration module is used to obtain the ambient sound detection signal and determine whether the current ambient sound meets the voiceprint registration conditions;

[0031] A first audio segment recording module, configured to record a first audio segment when the current ambient sound meets the voiceprint registration condition;

[0032] a first audio segment processing module, configured to process the first audio segment to obtain first voiceprint information and first feature information from the first audio segment;

[0033] A second audio segment recording module, configured to record a second audio segment when the first characteristic information reaches a preset decibel condition;

[0034] A second audio segment processing module, used for processing the second audio segment to obtain second voiceprint information and second feature information from the second audio segment;

[0035] The voiceprint feature matching module is used to extract the first voiceprint feature of the first audio segment and the second voiceprint feature of the second audio segment when the second feature information reaches the decibel preset condition, and calculate the matching degree between the first voiceprint feature and the second voiceprint feature, and store the audio segment that meets the preset matching degree as the user voiceprint in the voiceprint library.

[0036] In a third aspect, an embodiment of the present application provides a computer device, comprising a processor and a memory, wherein the memory stores a computer program, and the processor is used to execute the computer program to implement the voiceprint registration method of the above embodiment.

[0037] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, and when the computer program is executed on a processor, the voiceprint registration method of the above embodiment is implemented.

[0038] The embodiments of the present application have the following beneficial effects:

[0039] The present application provides a voiceprint registration method, device, computer equipment and storage medium that obtains an ambient sound detection signal to determine whether the current ambient sound meets the voiceprint registration conditions. When the current ambient sound meets the voiceprint registration conditions, the first audio segment is recorded to obtain the first voiceprint information and the first feature information. When the first feature information reaches the decibel preset condition, the second audio segment is recorded to obtain the second voiceprint information and the second feature information. When the second feature information reaches the decibel preset condition, the first voiceprint feature of the first audio segment and the second voiceprint feature of the second audio segment are extracted, and the matching degree of the first voiceprint feature and the second voiceprint feature is calculated, and the audio segment that meets the preset matching degree is stored in the voiceprint library as the user's voiceprint. The voiceprint registration method of the present application enables the audio segment stored in the voiceprint library to have a strong anti-interference ability, thereby ensuring the integrity, reliability and accuracy of the voiceprint registration, and finally obtaining a high-quality voiceprint. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 An application scenario diagram of a voiceprint registration method according to an embodiment of the present application is shown;

[0042] Figure 2 A flow chart of a voiceprint registration method according to an embodiment of the present application is shown;

[0043] Figure 3 Another flow chart of a voiceprint registration method according to an embodiment of the present application is shown;

[0044] Figure 4 Another flow chart of a voiceprint registration method according to an embodiment of the present application is shown;

[0045] Figure 5 Another flow chart of a voiceprint registration method according to an embodiment of the present application is shown;

[0046] Figure 6 A schematic diagram of the structure of a voiceprint registration device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0047] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0048] The components of the embodiments of the present application generally described and shown in the drawings herein may be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0049] Hereinafter, the terms "including", "having" and their cognates that can be used in various embodiments of the present application are intended only to indicate specific features, numbers, steps, operations, elements, components or a combination of the foregoing items, and should not be understood as first excluding the existence of one or more other features, numbers, steps, operations, elements, components or a combination of the foregoing items or increasing the possibility of one or more features, numbers, steps, operations, elements, components or a combination of the foregoing items. In addition, the terms "first", "second", "third" and the like are only used to distinguish descriptions and cannot be understood as indicating or implying relative importance.

[0050] Unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meanings as those generally understood by those skilled in the art to which the various embodiments of the present application belong. The terms (such as those defined in generally used dictionaries) will be interpreted as having the same meanings as the contextual meanings in the relevant technical field and will not be interpreted as having idealized meanings or overly formal meanings unless clearly defined in the various embodiments of the present application.

[0051] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0052] A voiceprint registration method of the present application takes into account the impact of different environments on user recorded audio, the impact of sudden noise or clipping on user recorded audio, etc., and proposes a voiceprint registration method. By obtaining the ambient sound detection signal, it is determined whether the current ambient sound meets the voiceprint registration conditions. When the current ambient sound meets the voiceprint registration conditions, the first audio segment is recorded to obtain the first voiceprint information and the first feature information. When the first feature information reaches the decibel preset condition, the second audio segment is recorded to obtain the second voiceprint information and the second feature information. When the second feature information reaches the decibel preset condition, the first voiceprint feature of the first audio segment and the second voiceprint feature of the second audio segment are extracted, and the matching degree of the first voiceprint feature and the second voiceprint feature is calculated, and the audio segment that meets the preset matching degree is stored in the voiceprint library as the user voiceprint. The voiceprint registration method of the present application records two audio segments, obtains the matching degree of the first voiceprint feature and the second voiceprint feature in the two audio segments, and stores the audio segment that meets the preset matching degree as the user voiceprint in the voiceprint library, so that the audio segment stored in the voiceprint library has a strong anti-interference ability, thereby ensuring the integrity, reliability and accuracy of the voiceprint registration, and finally obtaining a high-quality voiceprint.

[0053] A voiceprint registration method provided in an embodiment of the present application can be applied in Figure 1 In the application environment, specifically, the voiceprint registration method of the present application is applied in a computer system, the system includes: Figure 1 The client and server shown in the figure communicate with each other through the network. The client, also known as the user end, refers to the program corresponding to the server and providing local services to the client. The client can be installed on, but not limited to, various personal computers, laptops, smart phones, tablets, and portable wearable devices. The server can be implemented as an independent server or a server cluster consisting of multiple servers.

[0054] Figure 2 A flow chart of the voiceprint registration method of an embodiment of the present application is shown. The voiceprint registration method includes the following steps:

[0055] Step S100, obtaining an ambient sound detection signal, and determining whether the current ambient sound meets the voiceprint registration condition.

[0056] Specifically, when the user clicks to trigger "voiceprint registration", the system receives the ambient sound detection signal and issues an ambient detection prompt, reminding the user to keep quiet and entering a 3-second countdown. Then, after the countdown is completed, it determines whether the current ambient sound meets the voiceprint registration conditions. If so, the system uses a microphone or other sound collection device to capture the sound in the surrounding environment for a period of time. Otherwise, the system prompts "user keep quiet and start re-recording". The above "a period of time" usually refers to a shorter period of time, such as three seconds, five seconds, etc., which can be set according to actual needs.

[0057] After collecting the audio of the ambient sound sample, considering that the ambient sound sample may contain signals of various sound elements, such as human voice, ambient noise, background music, etc., a reasonable sound pressure threshold can be set to determine whether the current environment meets the voiceprint registration conditions. If the sound pressure of the ambient sound sample is lower than the threshold, it is considered that the current environment meets the voiceprint registration conditions and can proceed to the next step; otherwise, it needs to be re-recorded;

[0058] It is worth noting that the reasonable sound pressure threshold is set based on experimental data, aiming to ensure that voiceprint collection is performed in a sufficiently quiet environment. That is, only when the sound pressure decibel value of the current ambient sound is lower than the reasonable sound pressure threshold, the current environment is considered quiet enough and suitable for voiceprint registration.

[0059] Step S200: When the current ambient sound meets the voiceprint registration condition, record a first audio segment.

[0060] Specifically, whether voiceprint registration is possible is determined based on the sound pressure decibel value of the current ambient sound. If the sound pressure decibel value of the current ambient sound meets the voiceprint registration conditions, the system records the first audio segment when it receives a "start recording" signal triggered by the user. After the system receives an "end recording" signal triggered by the user, it ends recording the first audio segment, thereby obtaining the first audio segment.

[0061] For example, the sound pressure decibel value can be calculated by the following formula:

[0062]

[0063] Where: Lp represents the sound pressure of the current ambient sound collected; P represents the sound pressure actually measured in the current environment; P0 represents the reference sound pressure, and the value of P0 is 2×10 -5 Pa.

[0064] Step S300: Process the first audio segment to obtain first voiceprint information and first feature information from the first audio segment.

[0065] Specifically, the first audio segment is a segment of audio recorded after the current ambient sound meets the voiceprint registration conditions, and the first voiceprint information and the first feature information are information extracted from the first audio segment, wherein the first voiceprint information refers to the duration information of a segment of audio, and the first feature information refers to the signal-to-noise ratio information of a segment of audio.

[0066] For example, operations such as denoising, enhancing human voice, segmenting audio, filtering, gain control, and feature extraction may be performed on the first audio segment to extract the first voiceprint information and the first feature information.

[0067] Step S400: When the first characteristic information reaches a preset decibel condition, a second audio segment is recorded.

[0068] Specifically, the first characteristic information refers to the signal-to-noise ratio information of a segment of audio, and the decibel preset condition is the condition for triggering the recording of the second audio segment. When the system detects that the first characteristic information reaches the decibel preset condition, it starts recording the second segment of audio.

[0069] In this application, the decibel preset condition is 20dB, that is, when the first characteristic information is higher than 20dB, it jumps to the recording page of the second audio segment. Otherwise, the system prompts the user to "keep quiet during recording" and then re-record the first audio segment.

[0070] Step S500: Process the second audio segment to obtain second voiceprint information and second feature information from the second audio segment.

[0071] For example, the second audio segment can be processed by denoising, enhancing human voice, segmenting audio, filtering, gain control, feature extraction, etc. to obtain the second characteristic voiceprint information and second characteristic information in the second audio segment. The second voiceprint information is the duration information of the second audio segment, and the second characteristic information is the signal-to-noise ratio information of the second audio segment.

[0072] Step S600, when the second characteristic information reaches the decibel preset condition, extract the first voiceprint feature of the first audio segment and the second voiceprint feature of the second audio segment, calculate the matching degree between the first voiceprint feature and the second voiceprint feature, and store the audio segment that meets the preset matching degree as the user voiceprint in the voiceprint library.

[0073] For example, the matching degree can be obtained by using cosine similarity, Euclidean distance or other similarity metrics.

[0074] When the matching degree meets the preset matching degree, it is considered that the voiceprint consistency of the above two audio segments is high, the voiceprint registration is successful, and the voiceprint of the first audio segment is stored in the voiceprint library. Otherwise, the user is prompted that the voiceprint consistency of the two audio segments is too low, the voiceprint registration fails, and re-registration is required.

[0075] The voiceprint registration method of the embodiment of the present application records two audio segments, calculates the matching degree of the first voiceprint feature and the second voiceprint feature in the two audio segments, compares the matching degree with the preset matching degree, and when the matching degree exceeds the preset matching degree, it is determined that the voiceprints are consistent, and the first audio segment is stored in the voiceprint library as the user's voiceprint, so that the voiceprint features in the audio segment have strong anti-interference ability, thereby ensuring the integrity, reliability and accuracy of the voiceprint registration, and finally obtaining audio of better quality.

[0076] In an optional embodiment, if Figure 3 As shown, in step S300, before processing the first audio segment to obtain the first voiceprint information and the first feature information from the first audio segment, the method includes:

[0077] Step S301, obtaining a registration instruction from a user, and detecting the sound pressure decibel value of the current ambient sound according to the registration instruction.

[0078] Specifically, when the system receives the user's registration instruction, it performs ambient sound detection according to the registration instruction, and uses a microphone or other sound collection device to record short-term ambient sound. Then, it can calculate the sound pressure level (SPL) of the sound by analyzing the sound waveform captured by the microphone, sampling the sound signal, frequency analysis (such as FFT) and decibel conversion.

[0079] For example, the actual measured sound pressure and RMS are calculated by the following formula:

[0080]

[0081] Where: P represents the actual sound pressure measured in the current environment; A RMS represents the root mean square, x n Represents the value of the nth sample point, and N represents the total number of samples.

[0082] Step S302, determining whether the sound pressure decibel value of the current ambient sound reaches a preset recording condition.

[0083] Specifically, considering the background noise level in different environments, the system determines whether the environment is quiet enough by setting a reasonable recording preset condition. For example, if the decibel value of the sound pressure of the ambient sound detected exceeds the recording preset condition, the system may prompt the user to record in a quieter environment. If the decibel value of the sound pressure of the ambient sound detected does not exceed the recording preset condition, the system will prompt that recording is about to begin.

[0084] Step S303: When the sound pressure decibel value of the current ambient sound reaches the preset recording condition, the first audio segment is recorded.

[0085] Specifically, the recording preset conditions can be set according to actual needs and environmental characteristics. For example, if the registration process needs to be carried out in a relatively quiet environment, the recording preset conditions may be set to a lower level. On the contrary, if the registration process needs to be carried out in a relatively noisy environment, the recording preset conditions may be set to a higher level to ensure that background noise does not interfere with the recorded audio. In addition to the sound pressure decibel value, the recording preset conditions may also include other factors, such as the stability of the audio signal, the availability of the microphone device, etc.

[0086] The voiceprint registration method of the embodiment of the present application clarifies the conditions for recording the first audio segment, so as to obtain two recorded audio segments in the subsequent steps, and obtain the matching degree of the first voiceprint feature and the second voiceprint feature in the two audio segments, and store the audio segment that meets the preset matching degree as the user voiceprint in the voiceprint library, so that the audio segment stored in the voiceprint library has a strong anti-interference ability, thereby ensuring the integrity, reliability and accuracy of the voiceprint registration, and finally obtaining a high-quality voiceprint.

[0087] In an optional embodiment, if Figure 4 As shown, in step S300, the processing of the first audio segment to obtain first voiceprint information and first feature information from the first audio segment includes:

[0088] Step S304: preprocess the first audio segment, and input the first feature vector group obtained after the preprocessing into the neural network model for training to obtain first voiceprint information in the first audio segment, wherein the first voiceprint information includes first valid speech duration information and first invalid noise duration information.

[0089] Specifically, the original audio signal in the first audio segment is divided into frames, each frame usually lasting 25 milliseconds, each frame of audio is converted into a feature vector, and a first feature vector group is formed, and the first feature vector group is input into the VAD neural network model, and the VAD neural network model outputs a probability value, indicating the possibility that the frame belongs to the speech segment. Then, the probability value output by the model is compared with a preset threshold value, and if the probability output by the model is higher than the preset threshold value, the frame is marked as the first valid speech duration information, otherwise, it is marked as the first invalid speech duration information.

[0090] After obtaining the first valid speech duration information and the first invalid speech duration information, in order to reduce false detection and missed detection, the VAD neural network model usually undergoes some post-processing steps, such as de-jittering, smoothing filtering, etc., to ensure that the classification results of adjacent frames are more coherent and reduce short-term misclassification. Finally, the VAD neural network model generates a time series label, indicating which part is the valid speech duration information and which part is the invalid speech duration information.

[0091] Step S305: When the first effective voice duration information meets the first duration preset condition, first characteristic information is obtained.

[0092] Specifically, the first time preset condition is a preset time threshold, which is used to determine whether the voice uttered by the user is long enough for subsequent feature extraction or other processing. For example, the first time preset condition can be 25 seconds, 33 seconds, 35 seconds, etc. The time preset condition can vary according to different application scenarios.

[0093] The root mean square of the first valid speech duration information and the first invalid noise duration information is calculated to obtain the first feature information.

[0094] Specifically, if the first valid voice duration information meets the first duration preset condition, the root mean square of the first valid voice duration information and the first invalid voice duration information is calculated to obtain the first feature information; otherwise, if the first valid voice duration information does not meet the first duration preset condition, the system prompts "The duration of the voice is less than 30 seconds, and the first audio needs to be re-recorded."

[0095] The voiceprint registration method of the embodiment of the present application uses a VAD neural network model to obtain which part of the entire speech is valid voice duration information and which part is invalid voice duration information. For the subsequent matching degree, the matching degree is compared with the preset matching degree. When the matching degree exceeds the preset matching degree, it is determined that the voiceprints are consistent, and the first audio segment is stored in the voiceprint library as the user's voiceprint, so that the voiceprint features in the audio segment have strong anti-interference ability, thereby ensuring the integrity, reliability and accuracy of the voiceprint registration, and finally obtaining audio of better quality.

[0096] In an optional embodiment, if Figure 5 As shown, in step S500, the processing of the second audio segment to obtain second voiceprint information and second feature information from the second audio segment includes:

[0097] Step S501: When the first characteristic information meets the decibel preset condition, a second audio segment is recorded.

[0098] For example, the decibel preset condition can be set to 20dB. If the first characteristic information is higher than 20dB, the program will jump to the second audio segment recording page to start recording the second audio segment. Conversely, if the first characteristic information is lower than 20dB, the user will be prompted to "keep quiet during recording" and re-record the first audio segment.

[0099] Step S502, preprocessing the second audio segment, and inputting the second feature vector group obtained after the preprocessing into the neural network model for training to obtain second voiceprint information in the second audio segment, wherein the second voiceprint information includes second valid speech duration information and second invalid noise duration information;

[0100] Please refer to the above step S304, which will not be described in detail.

[0101] Step S503: When the second effective voice duration information meets the second duration preset condition, obtain second characteristic information.

[0102] Specifically, the second preset time condition is lower than the first preset time condition. For example, when the first preset time condition is 30 seconds, the second preset time condition can be set to 10 seconds. That is, when the second effective voice duration exceeds 10 seconds, it is deemed to meet the second preset time condition and obtain the second characteristic information.

[0103] Calculate the root mean square of the second valid speech duration information and the second invalid noise duration information to obtain the second feature information.

[0104] Specifically, the second feature information is obtained by analyzing the root mean square of the second valid speech duration information and the second invalid noise duration information.

[0105] The voiceprint registration method of the embodiment of the present application pre-processes the second audio segment, and inputs the second feature vector group obtained after the pre-processing into the neural network model for training to obtain the second voiceprint information in the second audio segment. In order to facilitate in the subsequent steps, by recording the matching degree of the first voiceprint feature and the second voiceprint feature in the two audio segments, the matching degree is compared with the preset matching degree. When the matching degree exceeds the preset matching degree, it is determined that the voiceprints are consistent, and the first audio segment is stored in the voiceprint library as the user's voiceprint, so that the voiceprint features in the audio segment have a strong anti-interference ability, thereby ensuring the integrity, reliability and accuracy of the voiceprint registration, and finally obtaining audio with better quality.

[0106] In one embodiment, extracting the first voiceprint feature of the first audio segment includes:

[0107] The first characteristic information in the first audio segment is detected to obtain the window sound pressure information of the first audio segment; the window sound pressure information is compared with the first invalid noise duration information to obtain the first voiceprint feature of the first audio segment.

[0108] Specifically, the information of the first feature information that is lower than the preset signal-to-noise ratio threshold is discarded, and the information of the part higher than the preset signal-to-noise ratio threshold is further processed, that is, the first effective speech duration information is finely divided to detect burst noise, and the sound pressure information in the window of the first audio segment is obtained by calculating the root mean square of the sound pressure in the window with a window length such as 1 second.

[0109] The sound pressure information in the window of the first audio segment is compared with the root mean square of the first invalid noise duration to determine the first voiceprint feature of the first audio segment. If it exceeds the preset signal-to-noise ratio threshold, it is identified as a sudden noise window, and the segment interfered by the sudden noise needs to be identified and removed in time.

[0110] Among them, based on the preset signal-to-noise ratio threshold, the signal-to-noise ratio threshold in the window is set to 2 to 3 times the preset value. That is, when the voice loudness in the window suddenly increases, it is considered to be a sudden noise, and the voice in the window needs to be discarded. In addition, the clips with clipping in the first valid voice duration information will also be removed. This ensures the purity of the voiceprint feature extraction. It reflects the strict control of the voiceprint data quality and enhances the accuracy of the voiceprint feature.

[0111] The voiceprint registration method of the embodiment of the present application ensures the purity of voiceprint feature extraction through the sound pressure information in the first audio segment window and the first voiceprint feature, reflects the strict control of voiceprint data quality, and enhances the accuracy of voiceprint features.

[0112] In one embodiment, extracting the second voiceprint feature of the second audio segment includes:

[0113] The second characteristic information in the second audio segment is detected to obtain the window sound pressure information of the second audio segment; the window sound pressure information is compared with the second invalid noise duration information to obtain the second voiceprint feature of the second audio segment.

[0114] With reference to the above steps of obtaining the window sound pressure information of the first audio segment and the first voiceprint feature, details are not repeated here.

[0115] The voiceprint registration method of the embodiment of the present application ensures the purity of voiceprint feature extraction through the sound pressure information in the second audio segment window and the second voiceprint feature, reflects the strict control of the voiceprint data quality, and enhances the accuracy of the voiceprint feature.

[0116] This application also proposes a voiceprint registration device, such as Figure 6 As shown, the voiceprint registration device includes:

[0117] The voiceprint registration module 61 is used to obtain an ambient sound detection signal and determine whether the current ambient sound meets the voiceprint registration condition.

[0118] The first audio segment recording module 62 is configured to record a first audio segment when the current ambient sound meets the voiceprint registration condition.

[0119] The first audio segment processing module 63 is used to process the first audio segment to obtain first voiceprint information and first feature information from the first audio segment.

[0120] The second audio segment recording module 64 is configured to record a second audio segment when the first characteristic information reaches a preset decibel condition.

[0121] The second audio segment processing module 65 is used to process the second audio segment to obtain second voiceprint information and second feature information from the second audio segment.

[0122] The voiceprint feature matching module 66 is used to extract the first voiceprint feature of the first audio segment and the second voiceprint feature of the second audio segment when the second feature information reaches the decibel preset condition, and calculate the matching degree between the first voiceprint feature and the second voiceprint feature, and store the audio segment that meets the preset matching degree as the user voiceprint in the voiceprint library.

[0123] Optionally, the voiceprint registration device further includes:

[0124] A sound pressure decibel value detection module is used to obtain a user's registration instruction and detect the sound pressure decibel value of the current ambient sound according to the registration instruction;

[0125] A sound pressure decibel value judgment module is used to judge whether the sound pressure decibel value of the current ambient sound meets the preset recording condition;

[0126] The first audio segment recording module is used to record the first audio segment when the sound pressure decibel value of the current ambient sound reaches the preset recording condition.

[0127] Optionally, the first audio segment processing module includes:

[0128] a first audio segment preprocessing unit, configured to preprocess the first audio segment, and input a first feature vector group obtained after the preprocessing into a neural network model for training to obtain first voiceprint information in the first audio segment, wherein the first voiceprint information includes first valid speech duration information and first invalid noise duration information;

[0129] A first characteristic information acquisition unit, configured to obtain first characteristic information when the first valid voice duration information meets a first duration preset condition;

[0130] When obtaining the first feature information, the first feature information acquisition unit obtains the first feature information by calculating the root mean square of the first valid speech duration information and the first invalid noise duration information.

[0131] Optionally, the second audio segment processing module includes:

[0132] A second audio segment recording unit, configured to record a second audio segment when the first characteristic information meets the decibel preset condition;

[0133] a second audio preprocessing unit, configured to preprocess the second audio segment, and input a second feature vector group obtained after the preprocessing into a neural network model for training to obtain second voiceprint information in the second audio segment, wherein the second voiceprint information includes second valid speech duration information and second invalid noise duration information;

[0134] A second characteristic information acquisition unit, configured to obtain second characteristic information when the second valid voice duration information meets a second duration preset condition;

[0135] When obtaining the second characteristic information, the second characteristic information obtaining unit obtains the second characteristic information by calculating the root mean square of the second valid speech duration information and the second invalid noise duration information.

[0136] Optionally, the first audio segment processing module further includes:

[0137] The first voiceprint feature acquisition unit is used to detect the first feature information in the first audio segment to obtain the window sound pressure information of the first audio segment; compare the window sound pressure information with the first invalid noise duration information to obtain the first voiceprint feature of the first audio segment.

[0138] Optionally, the second audio segment processing module further includes:

[0139] The second voiceprint feature acquisition unit is used to detect the second feature information in the second audio segment to obtain the sound pressure information within the window of the second audio segment; compare the sound pressure information within the window of the second audio segment with the second invalid noise duration information to obtain the second voiceprint feature of the second audio segment.

[0140] It can be understood that the device of this embodiment corresponds to a voiceprint registration method of the above embodiment, and the options in the above embodiment are also applicable to this embodiment, so they will not be described repeatedly here.

[0141] The present application also provides a computer device. Exemplarily, the computer device includes a processor and a memory, wherein the memory stores a computer program, and the processor runs the computer program to enable the terminal device to execute the functions of each module in the above-mentioned voiceprint registration method or the above-mentioned voiceprint registration device.

[0142] Among them, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including a central processing unit (CPU), a graphics processing unit (GPU) and a network processor (NP), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or at least one of other programmable logic devices, discrete gates or transistor logic devices, and discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc., which can implement or execute the disclosed methods, steps and logic block diagrams in the embodiments of the present application.

[0143] The memory may be, but is not limited to, a random access memory (RAM), a read only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), etc. The memory is used to store a computer program, and the processor may execute the computer program accordingly after receiving an execution instruction.

[0144] The present application also provides a computer-readable storage medium for storing the computer program used in the above-mentioned terminal device. For example, the computer-readable storage medium may include, but is not limited to, various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0145] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and structure diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in an alternative implementation, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the structure diagram and / or the flow diagram, and the combination of boxes in the structure diagram and / or the flow diagram, can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0146] In addition, the functional modules or units in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0147] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a smart phone, a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application.

[0148] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A voiceprint registration method, characterized in that: The voiceprint registration method comprises: Obtain the ambient sound detection signal to determine whether the current ambient sound meets the voiceprint registration conditions; When the current ambient sound meets the voiceprint registration condition, recording a first audio segment; Processing the first audio segment to obtain first voiceprint information and first feature information from the first audio segment includes: Preprocessing the first audio segment, and inputting a first feature vector group obtained after the preprocessing into a neural network model for training to obtain first voiceprint information in the first audio segment, wherein the first voiceprint information includes first valid speech duration information and first invalid noise duration information; When the first effective voice duration information meets the first duration preset condition, obtaining first characteristic information; The obtaining of the first characteristic information comprises: calculating the root mean square of the first valid speech duration information and the first invalid noise duration information to obtain the first characteristic information; When the first characteristic information reaches a preset decibel condition, recording a second audio segment; Processing the second audio segment to obtain second voiceprint information and second feature information from the second audio segment includes: When the first characteristic information meets the decibel preset condition, recording a second audio segment; Preprocessing the second audio segment, and inputting a second feature vector group obtained after the preprocessing into a neural network model for training to obtain second voiceprint information in the second audio segment, wherein the second voiceprint information includes second valid speech duration information and second invalid noise duration information; When the second effective voice duration information meets the second duration preset condition, obtaining second characteristic information; The obtaining of the second characteristic information comprises: calculating the root mean square of the second valid speech duration information and the second invalid noise duration information to obtain the second characteristic information; When the second characteristic information reaches the decibel preset condition, the first voiceprint feature of the first audio segment and the second voiceprint feature of the second audio segment are extracted, and the matching degree between the first voiceprint feature and the second voiceprint feature is calculated, and the audio segment that meets the preset matching degree is stored in the voiceprint library as the user voiceprint.

2. The voiceprint registration method according to claim 1, characterized in that: Before the processing of the first audio segment to obtain the first voiceprint information and the first feature information from the first audio segment, the method further includes: Obtaining a registration instruction from a user, and detecting the sound pressure decibel value of the current ambient sound according to the registration instruction; Determine whether the sound pressure decibel value of the current ambient sound meets the preset recording condition; When the sound pressure decibel value of the current ambient sound reaches the preset recording condition, the first audio segment is recorded.

3. The voiceprint registration method according to claim 1, characterized in that: Extracting a first voiceprint feature of the first audio segment includes: The first characteristic information in the first audio segment is detected to obtain the window sound pressure information of the first audio segment; the window sound pressure information is compared with the first invalid noise duration information to obtain the first voiceprint feature of the first audio segment.

4. The voiceprint registration method according to claim 1, characterized in that: Extracting a second voiceprint feature of the second audio segment includes: The second characteristic information in the second audio segment is detected to obtain the window sound pressure information of the second audio segment; the window sound pressure information of the second audio segment is compared with the second invalid noise duration information to obtain the second voiceprint feature of the second audio segment.

5. The voiceprint registration method according to claim 2, characterized in that: The sound pressure decibel value is calculated by the following formula: Where: Lp represents the sound pressure of the current ambient sound; P represents the sound pressure actually measured in the current environment; p0 represents the reference sound pressure, which is 2×10 -5 Pa; The actual measured sound pressure and root mean square are calculated by the following formula: Among them: A RMS represents the root mean square, P represents the actual sound pressure measured in the current environment; x n Represents the value of the nth sample point, and N represents the total number of samples.

6. A voiceprint registration device, characterized in that: The voiceprint registration device comprises: The voiceprint registration module is used to obtain the ambient sound detection signal and determine whether the current ambient sound meets the voiceprint registration conditions; A first audio segment recording module, configured to record a first audio segment when the current ambient sound meets the voiceprint registration condition; a first audio segment processing module, configured to process the first audio segment to obtain first voiceprint information and first feature information from the first audio segment; The first audio segment processing module comprises: a first audio segment preprocessing unit, configured to preprocess the first audio segment, and input a first feature vector group obtained after the preprocessing into a neural network model for training to obtain first voiceprint information in the first audio segment, wherein the first voiceprint information includes first valid speech duration information and first invalid noise duration information; A first characteristic information acquisition unit, configured to obtain first characteristic information when the first valid voice duration information meets a first duration preset condition; Wherein, when obtaining the first feature information, the first feature information obtaining unit obtains the first feature information by calculating the root mean square of the first valid voice duration information and the first invalid noise duration information; A second audio segment recording module, configured to record a second audio segment when the first characteristic information reaches a preset decibel condition; A second audio segment processing module, used for processing the second audio segment to obtain second voiceprint information and second feature information from the second audio segment; The second audio segment processing module comprises: A second audio segment recording unit, configured to record a second audio segment when the first characteristic information meets the decibel preset condition; a second audio preprocessing unit, configured to preprocess the second audio segment, and input a second feature vector group obtained after the preprocessing into a neural network model for training to obtain second voiceprint information in the second audio segment, wherein the second voiceprint information includes second valid speech duration information and second invalid noise duration information; A second characteristic information acquisition unit, configured to obtain second characteristic information when the second valid voice duration information meets a second duration preset condition; Wherein, when obtaining the second feature information, the second feature information obtaining unit obtains the second feature information by calculating the root mean square of the second valid voice duration information and the second invalid noise duration information; The voiceprint feature matching module is used to extract the first voiceprint feature of the first audio segment and the second voiceprint feature of the second audio segment when the second feature information reaches the decibel preset condition, and calculate the matching degree between the first voiceprint feature and the second voiceprint feature, and store the audio segment that meets the preset matching degree as the user voiceprint in the voiceprint library.

7. A computer device, characterized in that: The computer device comprises a processor and a memory, wherein the memory stores a computer program, and the processor is used to execute the computer program to implement the voiceprint registration method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: It stores a computer program, which, when executed on a processor, implements the voiceprint registration method according to any one of claims 1-5.

Images