Synthetic audio identification method and device, electronic equipment and storage medium
Through audio feature extraction and adaptive matching mechanisms, corresponding standard features are used for different audio quality, and the problem of insufficient accuracy and reliability of synthetic audio identification in the prior art is solved, achieving higher identification accuracy and reliability.
Patent Information
- Application Number
- CN202411504706.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-26
- Publication Date
- 2025-01-10
AI Technical Summary
The prior art is difficult to accurately and reliably identify synthetic audio, which may mislead the public and affect the authenticity and objectivity of news.
By extracting the audio feature, determining the corresponding standard features for matching based on the audio quality, using the adaptive processing mechanism, using more refined standard features for high-quality audio, and using low-quality standard features for low-quality audio to improve identification accuracy and reliability.
Under different audio quality conditions, relatively accurate and reliable identification results are achieved, misjudgment caused by quality differences is reduced, and the accuracy and reliability of synthetic audio identification is improved.
Smart Images
Figure CN119296569A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure belongs to the field of audio identification, and more specifically, relates to a synthetic audio identification method and device, electronic equipment, and storage medium. Background Art
[0002] With the development of audio technology, synthetic audio is becoming increasingly difficult to identify. Synthetic audio technology can generate highly realistic speech, which may be used to create false news, rumors or misleading information. For example, in the field of news media, audio news is an important form of information dissemination. If someone uses synthetic audio technology to forge the speech of news figures, it will mislead the public and cause social chaos. Identifying synthetic audio can ensure that the source of news audio is authentic and reliable, and maintain the authenticity and objectivity of the news.
[0003] Nowadays, the accuracy and reliability of audio identification are not high, and it is difficult to effectively identify whether the audio is synthetic audio.
[0004] It can be seen that an accurate and reliable synthetic audio identification method is urgently needed. Summary of the invention
[0005] The purpose of the present disclosure is to provide a synthetic audio identification method and device, electronic equipment, and storage medium to improve the accuracy and reliability of synthetic audio identification.
[0006] According to a first aspect of the present disclosure, a synthetic audio identification method is provided, comprising: Extracting features from the first audio to obtain a plurality of first features; In response to the audio quality of the first audio being greater than a first quality threshold, matching the plurality of first features with a plurality of first standard features to obtain an identification result; In response to the audio quality of the first audio being less than or equal to the first quality threshold, matching the plurality of first features with a plurality of second standard features to obtain an identification result; The audio qualities corresponding to the multiple first standard features and the multiple second standard features are different.
[0007] According to a second aspect of the embodiments of the present disclosure, there is provided a synthetic audio identification device, comprising: A feature extraction module, used to extract features from the first audio to obtain a plurality of first features; a first identification module, configured to match the plurality of first features with a plurality of first standard features to obtain an identification result in response to the audio quality of the first audio being greater than a first quality threshold; The second identification module is used to match the multiple first features with multiple second standard features to obtain the identification result in response to the audio quality of the first audio being less than or equal to the first quality threshold.
[0008] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of the above-mentioned synthetic audio identification method when executing the computer program.
[0009] According to a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned synthetic audio identification method are implemented.
[0010] The synthetic audio identification method and device, electronic device, and storage medium provided by the embodiments of the present disclosure have the following beneficial effects: The present disclosure determines which set of standard features to use for matching based on the audio quality of the first audio, and the above-mentioned adaptive processing mechanism ensures that relatively accurate and reliable identification results can be obtained under different audio quality conditions. For high-quality audio, using more sophisticated and stricter first standard features for matching can improve the accuracy and precision of identification; for low-quality audio, using low-quality audio features, i.e., second standard features, for matching can reduce misjudgments caused by poor quality. Since audios of different qualities differ in feature performance, using standard features that match the audio quality for matching can more effectively reduce identification errors caused by quality differences. The present disclosure improves the accuracy and reliability of identification results, so that the process has higher accuracy in applications such as audio classification and recognition, and improves the accuracy and reliability of synthetic audio identification. BRIEF DESCRIPTION OF THE DRAWINGS
[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0012] Figure 1 A schematic diagram of a flow chart of a synthetic audio identification method provided by an embodiment of the present disclosure; Figure 2 A structural block diagram of a synthetic audio identification device provided by an embodiment of the present disclosure; Figure 3 A schematic block diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION
[0013] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present disclosure. However, it should be clear to those skilled in the art that the present disclosure may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present disclosure with unnecessary details.
[0014] In order to make the purpose, technical solutions and advantages of the present disclosure more clear, specific embodiments will be described below in conjunction with the accompanying drawings.
[0015] Please refer to Figure 1 , Figure 1 A schematic diagram of a synthetic audio identification method provided in an embodiment of the present disclosure, the method comprising: S101: extracting features from a first audio to obtain a plurality of first features.
[0016] In this embodiment, the first audio may be music audio, voice audio, natural sound audio, etc., and the execution subject may obtain the first audio through wireless means such as Bluetooth or wired means such as data cables. The first audio may include more than one type of audio. For example, the first audio may be the audio of two people talking on the roadside, and the audio includes both the audio of the two people talking and the ambient audio.
[0017] Feature extraction is the process of extracting representative and distinguishing features from raw data. In audio processing, feature extraction can extract feature vectors that can describe the essential properties of audio from audio signals for subsequent analysis, classification, recognition and other tasks.
[0018] There may be multiple first features, and the first features may be time domain features, frequency domain features, or other features of the first audio, such as rhythm features, timbre features, and emotional features. The first features may be embodied in the form of numerical vectors, matrices, or structured data.
[0019] In this embodiment, features of the first audio may be extracted through a time domain analysis method, a frequency domain analysis method, a time-frequency analysis method, or a deep learning method.
[0020] S102: In response to the audio quality of the first audio being greater than a first quality threshold, matching the plurality of first features with the plurality of first standard features to obtain an identification result.
[0021] In this embodiment, the synthetic audio identification method further includes: The audio quality of the first audio is obtained by calculating the first audio based on at least one of a signal-to-noise ratio calculation method, a total harmonic distortion calculation method, an intermodulation distortion calculation method, and a frequency response method.
[0022] In this embodiment, the audio quality of the first audio can be calculated by signal-to-noise ratio calculation, total harmonic distortion measurement, feature extraction and classification, deep learning model or psychoacoustic model, etc. The audio quality can be considered comprehensively in terms of sound quality, loudness, sampling rate, bit rate, etc. For sound quality, its expression form can be signal-to-noise ratio, total harmonic distortion, etc. For loudness, its expression form can be decibel or sound pressure level, etc. For sampling rate, its expression form can be Hz. For bit rate, its expression form can be kbps.
[0023] All of the above aspects can be considered comprehensively or one or more of the most important indicators can be used as the audio quality evaluation criteria. For example, sound quality and bit rate are the core elements of audio quality. The sound quality of the first audio can be calculated separately, and the sound quality is 70dB (represented in the form of signal-to-noise ratio), and the bit rate of the first audio is 200kbps (kilobits per second).
[0024] The first quality threshold can be set manually. If there are two evaluation criteria in the above audio quality calculation process, then the corresponding first quality threshold should also have two corresponding thresholds to evaluate the audio quality of the first audio.
[0025] For example, when calculating the audio quality of the first audio, the sound quality and the bit rate of the first audio are calculated respectively, and then the corresponding first quality threshold should include the first sound quality threshold and the first bit rate threshold.
[0026] The above two evaluation criteria need to be satisfied at the same time and be greater than or equal to their corresponding thresholds before it can be determined that the audio quality of the first audio is greater than the first quality threshold.
[0027] The first feature is obtained by extracting features from the first audio. Since the features of different types of audio are different, for example, for human voice information, people will have pauses and changes in tone when speaking, etc., the voice information can be converted into text information, and the converted text information can be segmented based on the Chinese word segmentation of the Viterbi algorithm or the WordPiece algorithm, and the tone is consistent with the text based on the neural network model. The above-mentioned neural network model should be trained with a large amount of text information and the corresponding tone before use. For natural sounds, they may be chaotic, so the feature extraction methods corresponding to different types of audio should also be different.
[0028] The first standard feature is a feature that appears in various synthetic audios whose audio quality is higher than the first quality threshold, and it can be understood that the first standard feature is a feature of the synthetic audio. In this embodiment, matching can be performed by calculating the Euclidean distance or cosine similarity of multiple first features and multiple first standard features. Whether the matching is successful or not can be judged by matching degree, Euclidean distance or cosine similarity.
[0029] In this embodiment, each first feature can be traversed with all the first features to calculate the matching degree respectively, or the first standard features can be classified to obtain their corresponding classification information. When the first feature and the first standard feature are matched, only the first standard feature with the same classification information as the first feature can be calculated whether it matches the first feature.
[0030] When the match is successful, the first audio is determined to be a synthetic audio, and when the match is unsuccessful, the first audio is determined to be a natural audio.
[0031] S103: In response to the audio quality of the first audio being less than or equal to the first quality threshold, matching the plurality of first features with the plurality of second standard features to obtain an identification result; wherein the audio qualities corresponding to the plurality of first standard features and the plurality of second standard features are different.
[0032] In this embodiment, it is considered that low-quality audio may mask some features of the synthesized audio, thereby affecting the accuracy of identification. For example, a large amount of noise or distortion in the audio may interfere with the extraction and analysis of audio features by the identification algorithm. Therefore, the audio quality corresponding to the second standard feature in this embodiment is lower than the audio quality corresponding to the first standard feature, and the second standard feature is a feature of the synthesized audio with the same audio quality as the first audio.
[0033] A large number of high-quality synthetic audios and their corresponding first standard features can be collected. There are multiple first standard features to ensure that these audios cover different types and styles. The audio bit rate is reduced to the same bit rate as the first audio using audio processing tools or a certain step size. For example, the audio quality of the first audio is reflected in the form of bit rate. The bit rate of the first audio is 60kbps, and the first quality threshold is 192kbps. Since 60kbps is less than 192kbps, the audio quality of the first audio is too low, which will cause some synthetic features to be submerged. At this time, if it is still compared with the first standard feature, it will cause inaccurate or misjudgment of the identification result.
[0034] At this time, the bit rate of the music in the standard music library can be reduced to the audio quality of the first audio through an audio processing tool to obtain a new set of music libraries, and feature extraction is performed on these music to obtain a new set of features, which can be used as the second standard features. The above-mentioned standard music library is the source music library of the first standard feature, which contains various types of music in various fields, that is, the first standard feature is obtained by extracting features from the music in the standard music library.
[0035] Alternatively, the audio quality of the standard music can be adjusted through the first step length until the difference between the audio quality of the standard music and the first music is less than a first error threshold. The first error threshold can be set manually. It should be noted that the first step length is less than or equal to the first error threshold to avoid adjustment failure.
[0036] From the above, it can be concluded that the present disclosure determines the corresponding standard features for matching through the audio quality of the first audio, and the above-mentioned adaptive processing mechanism ensures that relatively accurate and reliable identification results can be obtained under different audio quality conditions. For high-quality audio, using a more refined and stricter first standard feature for matching can improve the accuracy and precision of identification; for low-quality audio, using low-quality audio features, namely the second standard features, for matching can reduce misjudgments caused by poor quality. Since audios of different qualities differ in feature performance, using standard features that match the audio quality for matching can more effectively reduce identification errors caused by quality differences. The present disclosure improves the accuracy and reliability of the identification results, so that the present disclosure has higher accuracy in applications such as audio classification and recognition.
[0037] In one embodiment of the present disclosure, feature extraction is performed on the first audio to obtain a plurality of first features, including: Separating the first audio based on sound source separation to obtain a plurality of second audios and classification information corresponding to the plurality of second audios; Determine a corresponding feature extraction method based on classification information corresponding to the plurality of second audios; Feature extraction is performed on the plurality of second audios based on a corresponding feature extraction method to obtain a plurality of first features.
[0038] In one embodiment of the present disclosure, before extracting features from the first audio to obtain a plurality of first features, the method further includes: An audio quality of the first audio is determined.
[0039] In this embodiment, the second audio is a different type or kind of music in the first audio, and the classification information refers to the different types or different sources of the second music. Since the features corresponding to different types of music are different, the corresponding feature extraction methods are also different. For example, the feature extraction method for human voice information can be Mel frequency cepstral coefficient detection, fundamental frequency change detection, spectrum and time domain analysis, etc.; the feature extraction method corresponding to natural sound can be feature extraction by spectrum analysis or time domain analysis.
[0040] In this embodiment, the first audio may be decomposed into multiple second audios and their corresponding classification information based on feature extraction and classification or deep learning.
[0041] For example, the first audio is a mixture of three types of sounds, and may be the audio of two people communicating outdoors in the rain accompanied by the sound of birds. The first audio may be decomposed into human voice audio, bird call audio, and rain audio based on feature extraction and classification, and the corresponding information is human voice, bird call, and rain sound, respectively. The feature extraction method corresponding to the human voice may be fundamental frequency change detection, and the fundamental frequency information may be extracted from the human voice. The statistical features such as the mean, variance, and dynamic range of the fundamental frequency, as well as the change curve of the fundamental frequency over time may be calculated, and matched with the first standard feature of the synthesized human voice in the first standard feature. The bird call audio may be subjected to time domain analysis to obtain the time domain features of the bird call audio, and matched with the first standard feature of the synthesized bird call in the first standard feature.
[0042] In this embodiment, the audio quality of the first audio can be calculated by signal-to-noise ratio calculation, total harmonic distortion measurement, feature extraction and classification, deep learning model or psychoacoustic model, that is, there can be one or more criteria for judging the audio quality of the first audio.
[0043] In this embodiment, before extracting features from the first audio, the following steps may also be included: Performing noise reduction processing on the first audio based on Fourier transform and noise elimination algorithm; Divide the first audio into a plurality of frames of duration; Multiple frames of duration are processed based on a window function.
[0044] In this embodiment, the noise elimination algorithm may be a Wiener filter algorithm, and the present disclosure may apply a Hamming window, a Hanning window, etc. to each frame to reduce the spectrum leakage effect.
[0045] It can be concluded from the above that the present disclosure decomposes the first audio into multiple second audios through sound source separation technology, and determines the corresponding feature extraction method based on classification information. It is possible to perform customized feature extraction for different types and kinds of music or sounds, so as to more accurately capture the essential attributes of each sound, which helps to improve the accuracy and efficiency of subsequent analysis, classification or recognition. This embodiment can process the first audio containing multiple types of sounds, and flexibly select the feature extraction method based on the classification information, so that the embodiment can be applied to a variety of scenarios, such as music classification, speech recognition, environmental sound monitoring, etc. At the same time, as the sound type increases or changes, the embodiment can also maintain its identification ability by updating the classification information and feature extraction method. The present disclosure can determine to match the first standard feature or compare with the second standard feature by evaluating the audio quality of the first audio, which helps to improve the accuracy and efficiency of the overall processing and improve the accuracy and reliability of synthetic audio identification.
[0046] In one embodiment of the present disclosure, before matching the plurality of first features with the plurality of second standard features, the method further includes: Get the first standard library; Obtaining a second standard library based on the audio quality of the first audio and the first standard library; The first standard feature is a feature in the first standard library; The second standard feature is a feature in the second standard library; The audio quality corresponding to the first standard library is higher than the audio quality corresponding to the second standard library.
[0047] In this embodiment, the first standard library contains various types of first standard features. The first standard library can be obtained by extracting features of various types of audio with audio quality higher than a first quality threshold. Multiple sub-standard libraries can be set in the first standard library, and audio types of the same type are stored in the same sub-standard library.
[0048] In response to the audio quality of the first audio being less than a first quality threshold, adjusting the audio quality corresponding to the first standard feature in the first standard library to the first audio quality; In response to the audio quality of the first audio being less than a first quality threshold, adjusting the audio quality corresponding to the first standard feature in the first standard library to an audio quality whose difference with the first audio quality is less than a first error threshold according to the first step, to obtain a plurality of second standard features; Among them, the audio quality corresponding to the second standard feature is less than the audio quality corresponding to the first standard feature, the audio quality corresponding to the second standard feature is equal to or the difference between the audio quality of the first audio and the first standard feature is less than the first error threshold, the first error threshold and the first step length can be set manually, but it should be noted that the first error threshold is greater than the first step length to avoid adjustment failure.
[0049] The second standard feature library may include multiple second standard sub-feature libraries, each type of feature corresponds to a second standard sub-feature library, and the same type of second standard features are stored in the same second standard sub-feature library.
[0050] It can be concluded from the above that the present disclosure obtains the first standard library, and then adjusts the second standard library according to the audio quality of the first audio, so that when performing feature matching, the second standard feature with a quality close to the first audio can be used, thereby ensuring the accuracy of feature matching. Matching using the second standard feature with a quality close to the first audio can reduce the matching error caused by the difference in audio quality. At the same time, the flexibility of the present disclosure is also improved, and it can adapt to inputs of different audio qualities. The present disclosure establishes a sub-standard library to make the feature library more orderly and efficient. The present disclosure stores the same type of features in the same sub-standard library, so that feature retrieval and matching can be convenient. At the same time, by adjusting the audio quality corresponding to the feature, the feature library can adapt to inputs of different audio qualities, thereby improving the utilization and applicability of the feature library, and improving the accuracy and reliability of synthetic audio identification.
[0051] In one embodiment of the present disclosure, a plurality of first features are matched with a plurality of first standard features to obtain an identification result, including: In response to the plurality of first features matching the plurality of first standard features, the identification result is a synthetic voice; In response to the plurality of first features not matching the plurality of first standard features, the identification result is a natural sound.
[0052] In one embodiment of the present disclosure, a plurality of first features are matched with a plurality of second standard features to obtain an identification result, including: In response to the plurality of first features matching the plurality of second standard features, the identification result is a synthetic voice; In response to the plurality of first features not matching the plurality of second standard features, the identification result is a natural voice.
[0053] In this embodiment, the first feature is matched with the first standard feature or the second standard feature, depending on the audio quality of the first audio. The first standard feature in the first standard library and the second standard feature in the second standard library are both features of the synthetic audio, and the difference is that the audio quality corresponding to the first standard feature in the first standard library is greater than the audio quality corresponding to the second standard feature in the second standard library.
[0054] Therefore, when the first feature matches the first standard feature or the second standard feature, the first feature is a feature of synthetic audio, and since the first feature is obtained by feature extraction of the first audio, the first audio is synthetic audio. Conversely, when the first feature does not match the first standard feature or the second standard feature, the first feature is a feature of natural audio, and the first audio is natural audio.
[0055] It should be noted that the above-mentioned natural audio refers to non-synthetic audio, not specifically to sounds in nature. Natural audio can include human voice audio, natural audio, and music audio, etc.
[0056] It can be concluded from the above that the present disclosure enhances the adaptability of the present disclosure to audio quality by introducing the first standard library and the second standard library, and selecting the matching standard according to the audio quality of the first audio, and can more flexibly process audio of different qualities, especially when processing low-quality audio, by adjusting the features in the first standard library to a quality level similar to that of the first audio (i.e., the second standard library), the accuracy and practicality of the matching are improved. The present disclosure improves the accuracy of identification by adjusting the features in the second standard library to match the audio quality of the first audio, reduces misjudgments caused by differences in audio quality, and improves the accuracy and robustness of identification. At the same time, the present disclosure further enhances the flexibility and adaptability of the present disclosure by setting multiple sub-standard libraries and error thresholds.
[0057] In one embodiment of the present disclosure, the synthetic audio identification method further includes: Segmenting the first audio to obtain a plurality of first segments; Obtaining a first identification result based on the plurality of first fragments; A final authentication result is obtained based on the first authentication result and the authentication result.
[0058] In this embodiment, the first segment is a segment in the first audio. It should be noted that there is no repeated segment in each first segment.
[0059] The segmentation can be performed based on the length of time, for example, the first audio is segmented into a segment every N seconds, or the first audio is sampled based on the same frequency to obtain multiple continuous first segments. It should be noted that in order to facilitate calculation, the lengths of the first segments need to be equal, and the lengths can be made consistent by truncation or zero padding.
[0060] The first identification result may be obtained by calculating the Euclidean distance or cosine similarity between two adjacent segments.
[0061] In this embodiment, the synthetic audio identification method further includes: in response to the Euclidean distance between the plurality of first segments being greater than or equal to a first distance threshold, the first identification result is the synthetic audio; In response to the Euclidean distance between the plurality of first segments being less than a first distance threshold, the first identification result is natural audio; The first distance threshold may be set according to actual needs.
[0062] In this embodiment, the synthetic audio identification method further includes: in response to the cosine similarity between the plurality of first segments being greater than or equal to a first similarity threshold, the first identification result is natural audio; In response to the cosine similarity between the plurality of first segments being less than a first similarity threshold, the first identification result is synthetic audio; The first similarity threshold can be set manually according to actual needs.
[0063] It should be noted that the first distance threshold is changed according to the vector value of the audio, and the first distance threshold corresponding to each type of audio should be different. The value range of the first similarity threshold should be The closer the cosine similarity value is to 1, the more similar the two vectors (audio signals) are; the closer the cosine similarity value is to -1, the less similar they are.
[0064] In this embodiment, the final identification result is obtained based on the first identification result and the identification result, including: In response to the first identification result and the identification result both being synthetic sounds, the final identification result being the synthetic sound; In response to one of the first identification result and the identification result being a synthetic voice, the final identification result is a possible synthetic voice; In response to the first identification result and the identification result both being natural sounds, the final identification result is a natural sound.
[0065] It can be concluded from the above that the present disclosure obtains multiple first segments by segmenting the first audio, and performs preliminary identification based on these segments (first identification result). Subsequently, the first identification result is combined with the identification result obtained based on the entire first audio to obtain a final identification result. It is possible to analyze the audio content more carefully and capture the subtle differences or inconsistencies that may exist in the audio, thereby improving the accuracy and reliability of the identification. In particular, when there is local synthesis or tampering in the audio, segmented identification can identify it more effectively. The present disclosure obtains the first identification result by calculating the Euclidean distance or cosine similarity between adjacent segments. This method is sensitive to the local features of the audio, and can reflect the coherence and consistency between audio segments, thereby enhancing the robustness of the present disclosure and being able to better cope with changes in various audio quality, formats and content. At the same time, by setting reasonable distance thresholds and similarity thresholds, it can flexibly adapt to different identification needs.
[0066] Corresponding to the synthetic audio identification method of the above embodiment, Figure 2 This is a structural block diagram of a synthetic audio identification device provided by an embodiment of the present disclosure. For ease of explanation, only the parts related to the embodiment of the present disclosure are shown. Figure 2 The synthetic audio identification device 20 includes: a feature extraction module 21, a first identification module 22 and a second identification module 23.
[0067] The feature extraction module 21 is used to extract features from the first audio to obtain a plurality of first features; A first identification module 22, configured to match the plurality of first features with the plurality of first standard features to obtain an identification result in response to the audio quality of the first audio being greater than a first quality threshold; The second identification module 23 is used to match the multiple first features with the multiple second standard features to obtain an identification result in response to the audio quality of the first audio being less than or equal to the first quality threshold.
[0068] In one embodiment of the present disclosure, the feature extraction module 21 is specifically configured to separate the first audio based on sound source separation to obtain a plurality of second audios and classification information corresponding to the plurality of second audios; Determine a corresponding feature extraction method based on classification information corresponding to the plurality of second audios; Feature extraction is performed on the plurality of second audios based on a corresponding feature extraction method to obtain a plurality of first features.
[0069] In one embodiment of the present disclosure, the synthetic audio identification device 20 further includes: an audio quality acquisition module; The audio quality acquisition module is used to determine the audio quality of the first audio.
[0070] In one embodiment of the present disclosure, the synthetic audio identification device 20 further includes: a standard library acquisition module; A standard library acquisition module, used to acquire a first standard library; Obtaining a second standard library based on the audio quality of the first audio and the first standard library; The first standard feature is a feature in the first standard library; The second standard feature is a feature in the second standard library; The audio quality corresponding to the first standard library is higher than the audio quality corresponding to the second standard library.
[0071] In one embodiment of the present disclosure, the first identification module 22 is specifically configured to, in response to the matching of the plurality of first features with the plurality of first standard features, identify the result as a synthetic sound; In response to the plurality of first features not matching the plurality of first standard features, the identification result is a natural sound.
[0072] In one embodiment of the present disclosure, the second identification module 23 is specifically configured to respond to the matching of the plurality of first features with the plurality of second standard features, and the identification result is a synthetic sound; In response to the plurality of first features not matching the plurality of second standard features, the identification result is a natural voice.
[0073] In one embodiment of the present disclosure, the synthetic audio identification device 20 further includes: an identification correction module; An identification and correction module, used for segmenting the first audio to obtain a plurality of first segments; Obtaining a first identification result based on the plurality of first fragments; A final authentication result is obtained based on the first authentication result and the authentication result.
[0074] In one embodiment of the present disclosure, the identification correction module is specifically configured to respond to the first identification result and the identification result being both synthetic sounds, and the final identification result being the synthetic sound; In response to one of the first identification result and the identification result being a synthetic voice, the final identification result is a possible synthetic voice; In response to the first identification result and the identification result both being natural sounds, the final identification result is a natural sound.
[0075] See also Figure 3 , Figure 3 A schematic block diagram of an electronic device provided by an embodiment of the present disclosure. Figure 3 The electronic device 300 in the embodiment shown may include: one or more processors 301, one or more input devices 302, one or more output devices 303 and one or more memories 304. The processors 301, input devices 302, output devices 303 and memories 304 communicate with each other via a communication bus 305. The memory 304 is used to store computer programs, which include program instructions. The processor 301 is used to execute the program instructions stored in the memory 304. The processor 301 is configured to call the program instructions to execute the functions of each module / unit in the above-mentioned device embodiments, such as Figure 2 The functions of modules 21 to 23 are shown.
[0076] It should be understood that in the embodiment of the present disclosure, the processor 301 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
[0077] The input device 302 may include a touch panel, a fingerprint collection sensor (for collecting the user's fingerprint information and fingerprint direction information), a microphone, etc., and the output device 303 may include a display (LCD, etc.), a speaker, etc.
[0078] The memory 304 may include a read-only memory and a random access memory, and provide instructions and data to the processor 301. A portion of the memory 304 may also include a non-volatile random access memory. For example, the memory 304 may also store information about the device type.
[0079] In a specific implementation, the processor 301, input device 302, and output device 303 described in the embodiments of the present disclosure can execute the implementation methods described in the first and second embodiments of the synthetic audio identification method provided in the embodiments of the present disclosure, and can also execute the implementation methods of the electronic device described in the embodiments of the present disclosure, which will not be repeated here.
[0080] In another embodiment of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by the processor, all or part of the processes in the above-mentioned embodiment method are implemented, and the computer program can also be completed by instructing the relevant hardware through the computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.
[0081] The computer-readable storage medium may be an internal storage unit of the electronic device of any of the aforementioned embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Furthermore, the computer-readable storage medium may also include both an internal storage unit of the electronic device and an external storage device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.
[0082] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this disclosure.
[0083] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the electronic devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0084] In the several embodiments provided in the present application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces or units, or it can be an electrical, mechanical or other form of connection.
[0085] 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 network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present disclosure.
[0086] In addition, each functional unit in each embodiment of the present disclosure 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.
[0087] The above are only specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present disclosure, and these modifications or replacements should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.
Claims
1. A synthetic audio identification method, characterized in that: include: Extracting features from the first audio to obtain a plurality of first features; In response to the audio quality of the first audio being greater than a first quality threshold, matching the plurality of first features with a plurality of first standard features to obtain an identification result; In response to the audio quality of the first audio being less than or equal to the first quality threshold, matching the plurality of first features with a plurality of second standard features to obtain an identification result; The audio qualities corresponding to the multiple first standard features and the multiple second standard features are different.
2. The synthetic audio identification method according to claim 1, characterized in that: The extracting features of the first audio to obtain a plurality of first features includes: Separating the first audio based on sound source separation to obtain a plurality of second audios and classification information corresponding to the plurality of second audios; Determine a corresponding feature extraction method based on classification information corresponding to the plurality of second audios; Based on the corresponding feature extraction method, feature extraction is performed on the multiple second audios to obtain the multiple first features.
3. The synthetic audio identification method according to claim 1, characterized in that: Before extracting features from the first audio to obtain a plurality of first features, the method further includes: An audio quality of the first audio is determined.
4. The synthetic audio identification method according to claim 1, characterized in that: Before matching the plurality of first features with the plurality of second standard features, the method further includes: Obtaining the first standard library; Obtaining the second standard library based on the audio quality of the first audio and the first standard library; The first standard feature is a feature in a first standard library; The second standard feature is a feature in a second standard library; The audio quality corresponding to the first standard library is higher than the audio quality corresponding to the second standard library.
5. The synthetic audio identification method according to claim 1, characterized in that: The matching of the plurality of first features with the plurality of first standard features to obtain an identification result includes: In response to the plurality of first features matching the plurality of first standard features, the identification result is a synthetic sound; In response to the plurality of first features not matching the plurality of first standard features, the identification result is a natural sound.
6. The synthetic audio identification method according to claim 1, characterized in that: The matching of the plurality of first features with the plurality of second standard features to obtain the identification result includes: In response to the plurality of first features matching the plurality of second standard features, the identification result is a synthetic sound; In response to the plurality of first features not matching the plurality of second standard features, the identification result is a natural sound.
7. The synthetic audio identification method according to claim 1, characterized in that: Also includes: Segmenting the first audio to obtain a plurality of first segments; Obtaining a first identification result based on the plurality of first fragments; A final identification result is obtained based on the first identification result and the identification result.
8. A synthetic audio identification device, characterized in that: include: A feature extraction module, used to extract features from the first audio to obtain a plurality of first features; a first identification module, configured to match the plurality of first features with a plurality of first standard features to obtain an identification result in response to the audio quality of the first audio being greater than a first quality threshold; The second identification module is used to match the multiple first features with multiple second standard features to obtain the identification result in response to the audio quality of the first audio being less than or equal to the first quality threshold.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.