Voiceprint signal updating method and device, electronic device and storage medium

Abstract

The present invention discloses a method for updating a voiceprint signal and a device, an electronic device and a storage medium thereof, and relates to the field of financial technology, wherein the updating method comprises: collecting an audio signal of a target user, and performing frame processing on the audio signal to obtain a voice frame set, extracting the voiceprint signal features of the audio signal based on the voice frame set, comparing the similarity between the voiceprint signal features of the audio signal and the voiceprint signal features of a pre-stored registered voiceprint, and when the similarity is greater than a preset similarity threshold, calculating the voice frame with the highest matching degree with the registered voiceprint to obtain a target voice frame, and using the target voice frame to replace the voice frame with the lowest audio signal in the registered voiceprint to update the audio signal of the registered voiceprint. The present invention solves the technical problem in the related art that the traditional identity authentication vector method cannot maintain the accuracy of voiceprint recognition in the long term.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and in particular to a method for updating a voiceprint signal, a device thereof, an electronic device and a storage medium. Background Art

[0002] Using voiceprint for identity authentication has become a common identification method. In related technologies, the traditional identity vector (i-vector) method is usually used for voiceprint modeling. This method can achieve good results under text-independent conditions. By mapping the speech into a feature space with category distinction, the similarity between speakers is directly compared based on the distance between features.

[0003] Voiceprints are mainly related to the physiological structure of the vocal organs and the way the vocal organs are manipulated. First, the physiological structure of the vocal organs (the position and size of the mouth, tongue, vocal cords, and oral cavity) will change dramatically as people age (the voices of children, teenagers, middle-aged people, and the elderly will change all the time), so the frequency range of the audio signal and the size of the sound tension will change accordingly. In addition, acquired learning will also cause the voiceprint to change (for example, new pronunciation habits learned after birth will change the previous pronunciation method). Therefore, when the audio signal of the person being identified changes with age or due to acquired learning, the existing i-vector method for voiceprint recognition will reduce the accuracy of recognition because it does not have strong robustness, especially in scenarios where voiceprint recognition is required for a long time. Tasks.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present invention provide a method for updating a voiceprint signal and a device thereof, an electronic device and a storage medium, so as to at least solve the technical problem in the related art that the traditional identity authentication vector method cannot maintain the accuracy of voiceprint recognition in the long term.

[0006] According to one aspect of an embodiment of the present invention, there is provided a method for updating a voiceprint signal, comprising: collecting an audio signal of a target user, and performing frame processing on the audio signal to obtain a voice frame set; extracting voiceprint signal features of the audio signal based on the voice frame set; comparing the similarity between the voiceprint signal features of the audio signal and the voiceprint signal features of a pre-stored registered voiceprint; when the similarity is greater than a preset similarity threshold, calculating a voice frame with the highest matching degree with the registered voiceprint to obtain a target voice frame; and replacing the voice frame with the lowest audio signal in the registered voiceprint with the target voice frame to update the audio signal of the registered voiceprint.

[0007] Optionally, after collecting the audio signal of the target user, the updating method further includes: converting the collected analog audio signal into a digital audio signal, and saving the converted digital audio signal in a preset storage format; performing energy normalization processing and pre-emphasis processing on the digital audio signal to obtain the audio signal to be framed, wherein the pre-emphasis processing refers to using a high-pass filter to emphasize the audio signal in the digital audio signal whose signal frequency is greater than a preset frequency threshold.

[0008] Optionally, after performing frame processing on the audio signal to obtain a speech frame set, the updating method further includes: performing windowing processing on the truncated signals between the speech frames in the speech frame set; after completing the windowing processing, calculating the signal short-time energy of each of the speech frames in the speech frame set; deleting the speech frames whose signal short-time energy is lower than a preset energy threshold, and retaining the speech frames whose signal short-time energy is greater than or equal to the preset energy threshold.

[0009] Optionally, based on the speech frame set, the step of extracting the voiceprint signal features of the audio signal includes: performing Fourier transform on each speech frame in the speech frame set to obtain a speech spectrum; performing a quadratic operation on the speech spectrum to obtain a speech energy spectrum; filtering the speech energy spectrum to obtain logarithmic energy; and performing discrete cosine transform on the logarithmic energy to obtain the voiceprint signal features of the audio signal.

[0010] Optionally, the step of filtering the speech energy spectrum to obtain logarithmic energy includes: after filtering the speech energy spectrum, determining multiple triangular filters distributed according to Mel frequency; and calculating the logarithmic energy of the triangular filter based on the frequency response parameters of the triangular filter.

[0011] Optionally, after comparing the similarity between the voiceprint signal feature of the audio signal and the voiceprint signal feature of a pre-stored registered voiceprint, the updating method further comprises: determining that the voice authentication of the target user is successful when the similarity is greater than a preset similarity threshold.

[0012] Optionally, after calculating the voice frame with the highest matching degree with the registered voiceprint and obtaining the target voice frame, the updating method also includes: obtaining the short-time energy of the voice signal of each frequency-divided voice frame in the registered voiceprint; and determining the voice frame corresponding to the minimum short-time energy of the voice signal as the voice frame with the lowest audio signal in the registered voiceprint.

[0013] According to another aspect of an embodiment of the present invention, there is also provided a device for updating a voiceprint signal, comprising: a framing unit for collecting an audio signal of a target user, and performing framing processing on the audio signal to obtain a speech frame set; an extraction unit for extracting the voiceprint signal features of the audio signal based on the speech frame set; a comparison unit for comparing the similarity between the voiceprint signal features of the audio signal and the voiceprint signal features of a pre-stored registered voiceprint; a calculation unit for calculating, when the similarity is greater than a preset similarity threshold, the voice frame with the highest matching degree with the registered voiceprint to obtain a target voice frame; and an updating unit for replacing the voice frame with the lowest audio signal in the registered voiceprint with the target voice frame to update the audio signal of the registered voiceprint.

[0014] Optionally, the updating device also includes: a first conversion module, used to convert the collected analog audio signal into a digital audio signal after collecting the audio signal of the target user, and save the converted digital audio signal in a preset storage format; a first processing module, used to perform energy normalization processing and pre-emphasis processing on the digital audio signal to obtain the audio signal to be framed, wherein the pre-emphasis processing refers to using a high-pass filter to emphasize the audio signal in the digital audio signal whose signal frequency is greater than a preset frequency threshold.

[0015] Optionally, the updating device also includes: a first windowing module, which is used to perform windowing processing on the truncated signals between the speech frames in the speech frame set after framing the audio signal to obtain a speech frame set; a first calculation module, which is used to calculate the signal short-time energy of each speech frame in the speech frame set after completing the windowing processing; and a first deletion module, which is used to delete the speech frames whose signal short-time energy is lower than a preset energy threshold, and retain the speech frames whose signal short-time energy is greater than or equal to the preset energy threshold.

[0016] Optionally, the extraction unit includes: a first transformation module, used to perform Fourier transform on each speech frame in the speech frame set to obtain a speech spectrum; a first operation module, used to perform a quadratic operation on the speech spectrum to obtain a speech energy spectrum; a first filtering module, used to filter the speech energy spectrum to obtain logarithmic energy; and a second transformation module, used to perform discrete cosine transform on the logarithmic energy to obtain the voiceprint signal characteristics of the audio signal.

[0017] Optionally, the first filtering module includes: a first determination submodule, used to determine multiple triangular filters distributed according to Mel frequency after filtering the speech energy spectrum; and a first calculation submodule, used to calculate the logarithmic energy of the triangular filter based on the frequency response parameters of the triangular filter.

[0018] Optionally, the updating device also includes: a first determination module, which is used to determine that the voice authentication of the target user is successful after comparing the similarity between the voiceprint signal characteristics of the audio signal and the voiceprint signal characteristics of a pre-stored registered voiceprint, if the similarity is greater than a preset similarity threshold.

[0019] Optionally, the updating device also includes: a first acquisition module, which is used to obtain the short-time energy of the voice signal of each frequency-divided voice frame in the registered voiceprint after calculating the voice frame with the highest matching degree with the registered voiceprint and obtaining the target voice frame; and a second determination module, which is used to determine the voice frame corresponding to the minimum voice signal short-time energy as the voice frame with the lowest audio signal in the registered voiceprint.

[0020] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is further provided, characterized in that the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned methods for updating the voiceprint signal.

[0021] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, characterized in that it comprises one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any one of the above-mentioned methods for updating the voiceprint signal.

[0022] In the present disclosure, the audio signal of the target user is collected, and the audio signal is processed by framing to obtain a speech frame set. Based on the speech frame set, the voiceprint signal features of the audio signal are extracted, and the similarity between the voiceprint signal features of the audio signal and the voiceprint signal features of the pre-stored registered voiceprint is compared. When the similarity is greater than a preset similarity threshold, the voice frame with the highest matching degree with the registered voiceprint is calculated to obtain the target voice frame, and the target voice frame is used to replace the voice frame with the lowest audio signal in the registered voiceprint to update the audio signal of the registered voiceprint. In the present disclosure, when the similarity is greater than the preset similarity threshold, it can be confirmed that the voiceprint authentication has passed. At this time, a certain voice frame of the user audio signal collected at the current moment can be used to replace the voice frame with the lowest audio signal in the original registered voiceprint, so that the user's authentication voiceprint can be updated in real time. This can not only overcome the difficulty of reducing the accuracy of voiceprint recognition due to changes in voiceprints of different age groups or changes in voiceprints due to acquired learning, but also effectively enhance the robustness of voiceprint recognition, thereby solving the technical problem that the traditional identity authentication vector method used in the related art cannot maintain the accuracy of voiceprint recognition in the long term. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0024] Figure 1 is a flow chart of an optional method for updating a voiceprint signal according to an embodiment of the present invention;

[0025] Figure 2 is a schematic diagram of an optional audio signal framing according to an embodiment of the present invention;

[0026] Figure 3 is a flowchart of an optional audio signal collection and preprocessing according to an embodiment of the present invention;

[0027] Figure 4 is a flow chart of an optional voiceprint signal feature extraction according to an embodiment of the present invention;

[0028] Figure 5 is a flow chart of an optional voiceprint recognition method based on anti-time variation according to an embodiment of the present invention;

[0029] Figure 6 is a flowchart of an optional voiceprint authentication method based on anti-time-variation according to an embodiment of the present invention;

[0030] Figure 7 is a schematic diagram of an optional device for updating a voiceprint signal according to an embodiment of the present invention;

[0031] Figure 8 The present invention is a hardware structure block diagram of an electronic device (or mobile device) for a method for updating a voiceprint signal according to an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] To facilitate those skilled in the art to understand the present invention, some terms or nouns involved in the embodiments of the present invention are explained below:

[0035] Mel-Frequency Cepstral Coefficients (MFCC): Mel frequency is proposed based on the auditory characteristics of the human ear. It has a nonlinear correspondence with the Hertz (Hz) frequency. The Mel-Frequency Cepstral Coefficients can use this relationship between them to calculate the Hz spectrum features. In this application, it can be used for voice data feature extraction and reducing the operation dimension.

[0036] Hamming window: It can be regarded as the sum of the spectra of three rectangular time windows, or the sum of three sin(t) type functions, and the two items in the brackets are moved to the left and right by π / T relative to the first spectrum window, so that the side lobes cancel each other out, eliminating high-frequency interference and leakage energy, and is suitable for non-periodic continuous signals.

[0037] Windowing: The main mathematical tool for digital signal processing is Fourier transform, which studies the relationship between the entire time domain and frequency domain. However, when using computers to implement engineering test signal processing, it is impossible to measure and calculate infinitely long signals. Instead, a finite time segment is taken for analysis. The method is to cut a time segment from the signal, and then use the cut signal time segment for periodic extension processing to obtain a virtual infinite-length signal. The signal can then be subjected to mathematical processing such as Fourier transform and correlation analysis.

[0038] It should be noted that the voiceprint signal updating method and device thereof in the present disclosure can be used in the financial technology field when updating the voiceprint signal, and can also be used in any field other than the financial technology field when updating the voiceprint signal. The present disclosure does not limit the application field of the voiceprint signal updating method and device thereof.

[0039] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0040] The following embodiments of the present invention can be applied to various systems / applications / devices for updating voiceprint signals. In the application scenarios of voiceprints, there are large differences in the voices of each person at different ages, and coupled with the influence of acquired learning, it is difficult to ensure the accuracy of existing voiceprint recognition. The present invention provides a voiceprint recognition and voiceprint update method based on an anti-time-varying algorithm, which can overcome the difficulty of low voiceprint recognition accuracy caused by different age groups and acquired learning, and effectively improve the recognition rate of various voiceprint application scenarios.

[0041] The present invention is described in detail below in conjunction with various embodiments.

[0042] Embodiment 1

[0043] According to an embodiment of the present invention, an embodiment of a method for updating a voiceprint signal is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0044] Figure 1 is a flow chart of an optional method for updating a voiceprint signal according to an embodiment of the present invention. Figure 1 As shown, the method comprises the following steps:

[0045] Step S101: collect the audio signal of the target user, and perform frame processing on the audio signal to obtain a speech frame set.

[0046] Step S102: extracting voiceprint signal features of the audio signal based on the speech frame set.

[0047] Step S103, comparing the similarity between the voiceprint signal feature of the audio signal and the voiceprint signal feature of the pre-stored registered voiceprint.

[0048] Step S104: when the similarity is greater than a preset similarity threshold, a speech frame having the highest matching degree with the registered voiceprint is calculated to obtain a target speech frame.

[0049] Step S105: The target speech frame is used to replace the speech frame with the lowest audio signal in the registered voiceprint to update the audio signal of the registered voiceprint.

[0050] Through the above steps, the audio signal of the target user can be collected, and the audio signal can be processed by framing to obtain a speech frame set. Based on the speech frame set, the voiceprint signal features of the audio signal are extracted, and the similarity between the voiceprint signal features of the audio signal and the voiceprint signal features of the pre-stored registered voiceprint is compared. When the similarity is greater than the preset similarity threshold, the voice frame with the highest matching degree with the registered voiceprint is calculated to obtain the target voice frame, and the target voice frame is used to replace the voice frame with the lowest audio signal in the registered voiceprint to update the audio signal of the registered voiceprint. In the embodiment of the present invention, when the similarity is greater than the preset similarity threshold, it can be confirmed that the voiceprint authentication has passed. At this time, a certain voice frame of the user audio signal collected at the current moment can be used to replace the voice frame with the lowest audio signal in the original registered voiceprint, so that the user's authentication voiceprint can be updated in real time. This can not only overcome the difficulty of reducing the accuracy of voiceprint recognition due to changes in voiceprints of different age groups or changes in voiceprints due to acquired learning, but also effectively enhance the robustness of voiceprint recognition, thereby solving the technical problem that the traditional identity authentication vector method used in the related art cannot maintain the accuracy of voiceprint recognition in the long term.

[0051] The embodiment of the present invention is described in detail below in combination with the above steps.

[0052] Step S101: collect the audio signal of the target user, and perform frame processing on the audio signal to obtain a speech frame set.

[0053] In an embodiment of the present invention, voiceprint audio files (including audio signals) can be collected by sound collection devices of various terminals (for example, mobile phones, telephones, self-service terminals, etc.). The general collection frequency can be 8 to 16kHz, and the quantization accuracy of 12bit and 16bit can be used. After collecting the audio signal of the target user, the collected audio signal can be framed to obtain a voice frame set.

[0054] Figure 2 is a schematic diagram of an optional audio signal framing according to an embodiment of the present invention, such as Figure 2 As shown, in order to better analyze the long speech, the stored long speech data needs to be divided into short speech segments of equal length (short speech 1, short speech 2, short speech 3, ..., short speech N, for example, a short speech segment of 10ms to 30ms can be divided into one frame), thereby obtaining a sound frame set.

[0055] Optionally, after collecting the audio signal of the target user, the updating method also includes: converting the collected analog audio signal into a digital audio signal, and saving the converted digital audio signal in a preset storage format; performing energy normalization and pre-emphasis processing on the digital audio signal to obtain an audio signal to be framed, wherein the pre-emphasis processing refers to using a high-pass filter to emphasize the audio signal in the digital audio signal whose signal frequency is greater than a preset frequency threshold.

[0056] In the embodiment of the present invention, the collected audio signal can firstly convert the analog audio signal into a digital audio signal using the Nyquist sampling theorem, and then save the converted digital audio signal in a preset storage format (eg, .wav format).

[0057] This embodiment can perform a series of preprocessing (including: energy normalization, pre-emphasis, framing, windowing and endpoint detection, etc.) on the digital audio signal to eliminate the deviation caused by the physical characteristics of the vocal organs and the external environment noise. The following is a schematic description of the preprocessing methods.

[0058] In this embodiment, energy normalization processing and pre-emphasis processing may be performed on the digital audio signal to obtain an audio signal to be framed, as follows:

[0059] The digital audio signal can be firstly processed for energy normalization to ensure that the volume of the voiceprint has little effect on the energy value of the speech frame. Then, the digital audio signal can be pre-emphasized. Since the speech signal is affected by the radiation of the mouth and nose and the excitation of the glottis, the high-frequency components above 800Hz will be attenuated at 6dB / octave. Therefore, pre-emphasis is to compensate for the energy loss in the high-frequency part. At the same time, the low-frequency part is attenuated to reduce the influence of noise, flatten the speech signal spectrum, and make the entire frequency band from low frequency to high frequency use the same signal-to-noise ratio to calculate the spectrum.

[0060] Pre-emphasis processing can use a first-order high-pass filter, the purpose of which is to emphasize the high-frequency part of the signal (that is, use a high-pass filter to emphasize the audio signal in the digital audio signal whose signal frequency is greater than a preset frequency threshold (which can be set according to actual conditions)), so that the original signal sampling transformation obtains a digital voice signal. In order to highlight the high-frequency component, it passes through the high-pass filter in turn, and the formula (1) can be used for pre-emphasis processing, where 0.9 is the pre-emphasis coefficient and Z represents a certain signal component.

[0061] H (Z) =1-0.9Z -1 (1);

[0062] After completing the energy normalization processing and the pre-emphasis processing, the frame division processing and the windowing processing can be performed.

[0063] Optionally, after the audio signal is framed to obtain a speech frame set, the updating method also includes: performing windowing processing on the truncated signals between the speech frames in the speech frame set; after completing the windowing processing, calculating the signal short-time energy of each speech frame in the speech frame set; deleting the speech frames whose signal short-time energy is lower than a preset energy threshold, and retaining the speech frames whose signal short-time energy is greater than or equal to the preset energy threshold.

[0064] In the embodiment of the present invention, after the audio signal is framed to obtain a speech frame set, the truncated signal between each speech frame in the speech frame set can be windowed. Since the frame processing will cause the truncation of the speech frame signal, thereby causing spectrum energy leakage, the signal at the truncation can be smoothly transitioned by windowing (the voiceprint speech signal can be processed by using a Hamming window), wherein the Hamming window time domain function is as shown in formula (2), wherein N is the frame length, which can generally be 512.

[0065]

[0066] After completing the windowing process, the signal short-time energy of each speech frame in the speech frame set can be calculated, and the speech frames whose signal short-time energy is lower than a preset energy threshold (which can be set according to actual conditions) are deleted, and the speech frames whose signal short-time energy is greater than or equal to the preset energy threshold are retained. In this embodiment, weak segments in the signal can be removed through endpoint detection, and valid speech segments can be retained. The signal short-time energy can be used to represent the strength of the speech, and the signal short-time energy E can be calculated according to the following formula (3): (n) , those with energy values ​​lower than a preset energy threshold (for example, 10) are eliminated, so that high-quality speech signals can be guaranteed, where X(n) represents a certain segment of speech signal, and it is assumed that there are N segments.

[0067]

[0068] After a series of preprocessing (including energy normalization, pre-emphasis, framing, windowing, and endpoint detection, etc.) in this embodiment, the audio signal will become a number of discrete speech frames to obtain a sound frame set.

[0069] Figure 3 is a flowchart of an optional audio signal collection and preprocessing according to an embodiment of the present invention, such as Figure 3 As shown, after collecting the audio signal (analog signal), the Nyquist sampling theorem can be used to convert the analog signal into a digital signal. After that, the digital signal is subjected to energy normalization, pre-emphasis, framing, windowing, and endpoint detection to obtain a set of audio frames.

[0070] Step S102: extracting voiceprint signal features of the audio signal based on the speech frame set.

[0071] Optionally, the step of extracting voiceprint signal features of an audio signal based on a speech frame set includes: performing Fourier transform on each speech frame in the speech frame set to obtain a speech spectrum; performing a quadratic operation on the speech spectrum to obtain a speech energy spectrum; filtering the speech energy spectrum to obtain logarithmic energy; and performing discrete cosine transform on the logarithmic energy to obtain voiceprint signal features of the audio signal.

[0072] Mel-frequency cepstral coefficient (MFCC) is a type of cepstral coefficient, which operates in the Mel-scale frequency domain and can accurately extract the nonlinear voiceprint signal characteristics of the human auditory system's perception frequency. The embodiment of the present invention can use the MFCC algorithm to extract the features of the audio signal.

[0073] In an embodiment of the present invention, Figure 4 is a flowchart of an optional voiceprint signal feature extraction according to an embodiment of the present invention. Figure 4 As shown, after obtaining the preprocessed audio signal, it can be multiplied by the Hamming window time domain function (formula (2)), and the signal can be converted from the time domain to the frequency domain. Then, the signal converted to the frequency domain (i.e., the voiceprint frequency domain signal) is Fourier transformed (i.e., each speech frame in the speech frame set is Fourier transformed, and the speech frame signal has been converted to a frequency domain signal). Formula (4) can be used for Fourier transform to obtain a speech spectrum, where N is the number of Fourier transform points, which can generally be 1024, and X(n) represents a frequency domain signal.

[0074]

[0075] Then, the speech spectrum is quadratically operated to obtain the speech energy spectrum, and then the speech energy spectrum is filtered to obtain the logarithmic energy S(m). After that, the logarithmic energy S(m) is discrete cosine transformed to obtain the voiceprint signal feature of the audio signal. The logarithmic energy S(m) can be discrete cosine transformed using formula (5) to obtain the voiceprint signal feature, where S(m) is the mth logarithmic energy, M is the number of filters used to filter the logarithmic energy S(m), and n is the number of voiceprint signal features.

[0076]

[0077] Optionally, the step of filtering the speech energy spectrum to obtain logarithmic energy includes: after filtering the speech energy spectrum, determining multiple triangular filters distributed according to Mel frequency; and calculating the logarithmic energy of the triangular filter based on the frequency response parameters of the triangular filter.

[0078] In the embodiment of the present invention, after filtering the energy spectrum (for example, Mel filtering), a plurality of triangular filters distributed according to the Mel frequency can be determined, and then the logarithmic energy S(m) of these filters is calculated according to formula (6) through these triangular filters (that is, the logarithmic energy of the triangular filter is calculated based on the frequency response parameters of the triangular filter), where H m (k) is the frequency response of the filter, and M is the number of filters, which can generally be 24.

[0079]

[0080] Step S103, comparing the similarity between the voiceprint signal feature of the audio signal and the voiceprint signal feature of the pre-stored registered voiceprint.

[0081] Optionally, after comparing the similarity between the voiceprint signal feature of the audio signal and the voiceprint signal feature of the pre-stored registered voiceprint, the updating method further includes: when the similarity is greater than a preset similarity threshold, determining that the voice authentication of the target user is successful.

[0082] In an embodiment of the present invention, voiceprint signal feature extraction can be performed on the pre-stored registered voiceprint. After the voiceprint signal feature of the audio signal in the most recent time is extracted, the similarity of the voiceprint signal features between the two can be compared. When the similarity is less than or equal to a preset similarity threshold (which can be set according to actual conditions), it can be determined that the voiceprint is not the person himself; when the similarity is greater than the preset similarity threshold, it is determined that the voice authentication of the target user is successful, that is, the voiceprint is the person's voiceprint, and the current user to be authenticated and the user to whom the registered voiceprint belongs are the same person.

[0083] Step S104: when the similarity is greater than a preset similarity threshold, a speech frame having the highest matching degree with the registered voiceprint is calculated to obtain a target speech frame.

[0084] In an embodiment of the present invention, when the similarity is greater than a preset similarity threshold, the short-time energy of the speech signal of several pre-processed frequency-divided speech frames in the most recently collected voiceprint can be calculated. After comparison, the speech frame with the highest match with the registered voiceprint can be selected to obtain the target speech frame.

[0085] Optionally, after calculating the voice frame with the highest matching degree with the registered voiceprint and obtaining the target voice frame, the updating method also includes: obtaining the short-time energy of the voice signal of each frequency-divided voice frame in the registered voiceprint; and determining the voice frame corresponding to the minimum short-time energy of the voice signal as the voice frame with the lowest audio signal in the registered voiceprint.

[0086] In an embodiment of the present invention, after obtaining the speech frame with the highest matching degree with the registered voiceprint, the short-time energy of the speech signal of each frequency-divided speech frame in the registered voiceprint can be calculated, and the speech frame corresponding to the smallest short-time energy of the speech signal can be determined as the speech frame with the lowest audio signal in the registered voiceprint.

[0087] Step S105: The target speech frame is used to replace the speech frame with the lowest audio signal in the registered voiceprint to update the audio signal of the registered voiceprint.

[0088] In an embodiment of the present invention, after determining the voice frame with the highest match degree with the registered voiceprint and the voice frame with the lowest audio signal in the registered voiceprint, the voice frame with the highest match degree with the registered voiceprint (i.e., the target voice frame) can replace the voice frame with the lowest audio signal in the registered voiceprint to update the audio signal of the registered voiceprint.

[0089] This embodiment can always use the voice frame with the highest matching degree with the registered voiceprint in the latest voiceprint to replace the voice frame with the lowest audio signal in the registered voiceprint, so that the user's voiceprint information is continuously updated to ensure the accuracy of voiceprint recognition.

[0090] Figure 5 is a flow chart of an optional voiceprint recognition method based on time-varying resistance according to an embodiment of the present invention, such as Figure 5 As shown, a voiceprint of a person registered at time A (i.e., the previous time or a self-selected time in the historical process) is obtained, and the voiceprint of the person registered and identified at time A is obtained. After audio signal preprocessing, it is framed into several short voice frames (including: short voice 1, short voice 2, short voice 3, ..., short voice N). When voiceprint authentication is performed at the next time B, the voiceprint of the person registered and identified at time B can be obtained. It is also first preprocessed with an audio signal, and then framed into several short voice frames (including: short voice 1, short voice 2, short voice 3, ..., short voice N). After the MFCC algorithm is used to extract feature parameters, it is determined whether the similarity is greater than the set threshold. If the similarity is less than the set threshold, it is determined to be not the person's voice; if the similarity is greater than the set threshold, it is recognized as the person's voiceprint voice, and the scores of the preprocessed short voice signal frames are calculated (that is, the short voice score of each short voice signal frame is calculated, including : short speech 1 score, short speech 2 score, short speech 3 score, ..., short speech N score), compare and select the short speech frame with the highest match with the voiceprint of the person registered for recognition at time A, replace the speech frame with the weakest signal in the voiceprint registered at time A (that is, by calculating the short-time energy of the speech signal, select the short speech with the lowest short-time energy), and form the latest registered voiceprint audio signal at time B (that is, short speech 1, short speech 2, short speech 3, ..., the latest short speech N can be obtained, and the latest short speech N is the short speech frame with the highest match with the voiceprint of the person registered for recognition at time A in the replaced voiceprint of the person registered for recognition at time B).

[0091] Since voiceprint registration has more stringent requirements on the length of voiceprints, if the registered voiceprint is short, it will affect the registered voiceprint and the subsequent recognition effect. Therefore, when customers register voiceprints, they are required to read text content for more than a preset time (for example, 20 seconds) and multiple times (for example, more than three times). For the training process, it is necessary to extract the speaker's voice features for training and learning, thereby establishing a voice model library. However, in the subsequent authentication process, in order to ensure the customer experience, it is often only necessary to read a text file with a lower preset time (for example, less than 10 seconds) and read it once for recognition. Therefore, in order to ensure that the recognition rate of voiceprints does not decrease with time, the customer's recognition process can be used to update the latest registered voiceprint audio signal.

[0092] Figure 6 is a flowchart of an optional voiceprint authentication method based on time-varying resistance according to an embodiment of the present invention. Figure 6 As shown, the voiceprint can be registered at time A. When voiceprint authentication is performed at time B, it is determined whether the authentication is passed. If the authentication is passed, a short voice segment in the voiceprint signal authenticated at time B can be used to update the voiceprint library registered at time A (i.e., time B replaces the voiceprint segment at time A to update the latest voiceprint library), otherwise the authentication fails. Afterwards, when voiceprint authentication is performed at time C, it is determined whether the authentication is passed. If the authentication is passed, a short voice segment in the voiceprint signal authenticated at time C can be used to update the voiceprint library registered at time B (i.e., time C replaces the voiceprint segment at time B to update the latest voiceprint library), otherwise the authentication fails. Similarly, when voiceprint authentication is performed at time N, it is determined whether the authentication is passed. If the authentication is passed, a short voice segment in the voiceprint signal authenticated at time N can be used to update the voiceprint library registered at time N-1 (i.e., time N replaces the voiceprint segment at time N-1 to update the latest voiceprint library), otherwise the authentication fails. In this way, it can be ensured that the voiceprint library is always the latest and highest quality voiceprint library, which effectively improves the recognition efficiency and accuracy each time.

[0093] In the embodiment of the present invention, the above-mentioned voiceprint updating method can be used to update the previously registered voiceprint with the latest voiceprint, so as to keep the voiceprint library always the latest and highest quality voiceprint library, which can not only overcome the difficulty of low voiceprint recognition accuracy caused by different age groups and acquired learning, but also effectively improve the recognition rate of various voiceprint application scenarios.

[0094] Embodiment 2

[0095] The voiceprint signal updating device provided in this embodiment includes a plurality of implementation units, each of which corresponds to each implementation step in the above-mentioned first embodiment.

[0096] Figure 7is a schematic diagram of an optional device for updating a voiceprint signal according to an embodiment of the present invention, such as Figure 7 As shown, the updating device may include: a framing unit 70, an extracting unit 71, a comparing unit 72, a calculating unit 73, and an updating unit 74, wherein:

[0097] The framing unit 70 is used to collect the audio signal of the target user and perform framing processing on the audio signal to obtain a speech frame set;

[0098] An extraction unit 71, configured to extract a voiceprint signal feature of an audio signal based on a speech frame set;

[0099] A comparing unit 72, configured to compare the similarity between the voiceprint signal feature of the audio signal and the voiceprint signal feature of the pre-stored registered voiceprint;

[0100] A calculation unit 73 is used to calculate the voice frame with the highest matching degree with the registered voiceprint when the similarity is greater than a preset similarity threshold, and obtain a target voice frame;

[0101] The updating unit 74 is configured to replace the speech frame with the lowest audio signal in the registered voiceprint with the target speech frame to update the audio signal of the registered voiceprint.

[0102] The above-mentioned updating device can collect the audio signal of the target user through the framing unit 70, and perform framing processing on the audio signal to obtain a speech frame set, extract the voiceprint signal characteristics of the audio signal based on the speech frame set through the extraction unit 71, compare the similarity between the voiceprint signal characteristics of the audio signal and the voiceprint signal characteristics of the pre-stored registered voiceprint through the comparison unit 72, and calculate the speech frame with the highest matching degree with the registered voiceprint through the calculation unit 73 when the similarity is greater than the preset similarity threshold to obtain the target speech frame, and use the target speech frame to replace the speech frame with the lowest audio signal in the registered voiceprint through the updating unit 74 to update the audio signal of the registered voiceprint. In an embodiment of the present invention, when the similarity is greater than a preset similarity threshold, it can be confirmed that the voiceprint authentication has passed. At this time, a voice frame of the user audio signal collected at the current moment can replace the voice frame with the lowest audio signal in the original registered voiceprint, so that the user's authentication voiceprint can be updated in real time. This not only overcomes the difficulty of reducing the accuracy of voiceprint recognition due to changes in voiceprints of different age groups or due to changes in voiceprints due to acquired learning, but also effectively enhances the robustness of voiceprint recognition, thereby solving the technical problem in the related technology that the traditional identity authentication vector method cannot maintain the accuracy of voiceprint recognition in the long term.

[0103] Optionally, the updating device also includes: a first conversion module, used to convert the collected analog audio signal into a digital audio signal after collecting the audio signal of the target user, and save the converted digital audio signal in a preset storage format; a first processing module, used to perform energy normalization and pre-emphasis processing on the digital audio signal to obtain the audio signal to be framed, wherein the pre-emphasis processing refers to the use of a high-pass filter to emphasize the audio signal in the digital audio signal whose signal frequency is greater than a preset frequency threshold.

[0104] Optionally, the updating device also includes: a first windowing module, which is used to perform windowing processing on the truncated signals between each speech frame in the speech frame set after framing the audio signal to obtain a speech frame set; a first calculation module, which is used to calculate the signal short-time energy of each speech frame in the speech frame set after completing the windowing processing; and a first deletion module, which is used to delete the speech frames whose signal short-time energy is lower than a preset energy threshold, and retain the speech frames whose signal short-time energy is greater than or equal to the preset energy threshold.

[0105] Optionally, the extraction unit includes: a first transformation module, used to perform Fourier transform on each speech frame in the speech frame set to obtain a speech spectrum; a first operation module, used to perform a quadratic operation on the speech spectrum to obtain a speech energy spectrum; a first filtering module, used to filter the speech energy spectrum to obtain logarithmic energy; and a second transformation module, used to perform discrete cosine transform on the logarithmic energy to obtain voiceprint signal characteristics of the audio signal.

[0106] Optionally, the first filtering module includes: a first determination submodule, used to determine multiple triangular filters distributed according to Mel frequency after filtering the speech energy spectrum; and a first calculation submodule, used to calculate the logarithmic energy of the triangular filter based on the frequency response parameters of the triangular filter.

[0107] Optionally, the updating device also includes: a first determination module, which is used to determine that the voice authentication of the target user is successful after comparing the similarity between the voiceprint signal characteristics of the audio signal and the voiceprint signal characteristics of the pre-stored registered voiceprint, if the similarity is greater than a preset similarity threshold.

[0108] Optionally, the updating device also includes: a first acquisition module, which is used to obtain the short-time energy of the voice signal of each frequency-divided voice frame in the registered voiceprint after calculating the voice frame with the highest matching degree with the registered voiceprint and obtaining the target voice frame; a second determination module, which is used to determine the voice frame corresponding to the minimum voice signal short-time energy as the voice frame with the lowest audio signal in the registered voiceprint.

[0109] The above-mentioned updating device may also include a processor and a memory. The above-mentioned framing unit 70, extraction unit 71, comparison unit 72, calculation unit 73, updating unit 74, etc. are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0110] The processor includes a kernel, which retrieves the corresponding program unit from the memory. One or more kernels can be set, and the audio signal of the registered voiceprint can be updated by adjusting the kernel parameters.

[0111] The above-mentioned memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one storage chip.

[0112] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that is initialized with the following method steps: collecting an audio signal of a target user, and performing frame processing on the audio signal to obtain a speech frame set, extracting voiceprint signal features of the audio signal based on the speech frame set, and comparing the similarity between the voiceprint signal features of the audio signal and the voiceprint signal features of a pre-stored registered voiceprint, and when the similarity is greater than a preset similarity threshold, calculating the voice frame with the highest match with the registered voiceprint to obtain a target voice frame, and using the target voice frame to replace the voice frame with the lowest audio signal in the registered voiceprint to update the audio signal of the registered voiceprint.

[0113] According to another aspect of an embodiment of the present invention, a computer-readable storage medium is provided, characterized in that the computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned methods for updating the voiceprint signal.

[0114] According to another aspect of an embodiment of the present invention, there is further provided an electronic device, characterized in that it comprises one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors implement any one of the above-mentioned methods for updating the voiceprint signal.

[0115] Figure 8 FIG. 1 is a hardware structure block diagram of an electronic device (or mobile device) for a method for updating a voiceprint signal according to an embodiment of the present invention. Figure 8As shown, the electronic device may include one or more (102a, 102b, ..., 102n are used to illustrate) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), and a memory 104 for storing data. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a keyboard, a power supply and / or a camera. It can be understood by those skilled in the art that Figure 8 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 8 More or fewer components as shown, or with Figure 8 Different configurations are shown.

[0116] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0117] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0118] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0119] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0120] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0121] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

[0122] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for updating a voiceprint signal, characterized in that: include: Collecting an audio signal of a target user and performing frame processing on the audio signal to obtain a speech frame set; Extracting voiceprint signal features of the audio signal based on the speech frame set; Comparing the similarity between the voiceprint signal feature of the audio signal and the voiceprint signal feature of the pre-stored registered voiceprint; When the similarity is greater than a preset similarity threshold, a voice frame having the highest matching degree with the registered voiceprint is calculated to obtain a target voice frame; The target voice frame is used to replace the voice frame with the lowest audio signal in the registered voiceprint, so as to update the audio signal of the registered voiceprint, wherein the audio signal in the registered voiceprint is a short voice frame after frame processing; After calculating the voice frame with the highest matching degree with the registered voiceprint and obtaining the target voice frame, the updating method also includes: obtaining the short-time energy of the voice signal of each frequency-divided voice frame in the registered voiceprint; and determining the voice frame corresponding to the minimum short-time energy of the voice signal as the voice frame with the lowest audio signal in the registered voiceprint.

2. The updating method according to claim 1, characterized in that: After collecting the audio signal of the target user, the updating method further includes: Converting the collected analog audio signal into a digital audio signal, and saving the converted digital audio signal in a preset storage format; The digital audio signal is subjected to energy normalization processing and pre-emphasis processing to obtain the audio signal to be framed, wherein the pre-emphasis processing refers to using a high-pass filter to perform emphasis processing on the audio signal whose signal frequency in the digital audio signal is greater than a preset frequency threshold.

3. The updating method according to claim 1, characterized in that: After the audio signal is subjected to frame processing to obtain a speech frame set, the updating method further comprises: Performing windowing processing on truncated signals between speech frames in the speech frame set; After completing the windowing process, calculating the signal short-time energy of each of the speech frames in the speech frame set; The speech frames whose signal short-time energy is lower than a preset energy threshold are deleted, and the speech frames whose signal short-time energy is greater than or equal to the preset energy threshold are retained.

4. The updating method according to claim 1, characterized in that: The step of extracting the voiceprint signal feature of the audio signal based on the speech frame set includes: Performing Fourier transform on each speech frame in the speech frame set to obtain a speech spectrum; Performing a quadratic operation on the speech spectrum to obtain a speech energy spectrum; Performing filtering processing on the speech energy spectrum to obtain logarithmic energy; Performing discrete cosine transform on the logarithmic energy to obtain a voiceprint signal feature of the audio signal.

5. The updating method according to claim 4, characterized in that: The step of filtering the speech energy spectrum to obtain logarithmic energy includes: After filtering the speech energy spectrum, determining a plurality of triangular filters distributed according to Mel frequencies; The logarithmic energy of the triangular filter is calculated based on the frequency response parameters of the triangular filter.

6. The updating method according to claim 1, characterized in that: After comparing the similarity between the voiceprint signal feature of the audio signal and the voiceprint signal feature of the pre-stored registered voiceprint, the updating method further includes: When the similarity is greater than a preset similarity threshold, it is determined that the voice authentication of the target user is successful.

7. A device for updating a voiceprint signal, characterized in that: include: A framing unit, used to collect an audio signal of a target user and perform framing processing on the audio signal to obtain a speech frame set; An extraction unit, configured to extract a voiceprint signal feature of the audio signal based on the speech frame set; A comparing unit, configured to compare the similarity between the voiceprint signal feature of the audio signal and the voiceprint signal feature of a pre-stored registered voiceprint; A calculation unit, configured to calculate a speech frame having the highest matching degree with the registered voiceprint when the similarity is greater than a preset similarity threshold, and obtain a target speech frame; An updating unit, configured to replace the speech frame with the lowest audio signal in the registered voiceprint with the target speech frame, so as to update the audio signal of the registered voiceprint, wherein the audio signal in the registered voiceprint is a short speech frame after frame processing; The updating device also includes: a first acquisition module, which is used to obtain the short-time energy of the voice signal of each frequency-divided voice frame in the registered voiceprint after calculating the voice frame with the highest matching degree with the registered voiceprint and obtaining the target voice frame; and a second determination module, which is used to determine the voice frame corresponding to the minimum short-time energy of the voice signal as the voice frame with the lowest audio signal in the registered voiceprint.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the method for updating the voiceprint signal according to any one of claims 1 to 6.

9. An electronic device, characterized in that: The invention comprises one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method for updating the voiceprint signal as described in any one of claims 1 to 6.

Images