Audio slicing processing method, system and storage medium for audio feature analysis

Abstract

The present invention provides an audio slicing processing method, system and storage medium for audio feature analysis, which relates to the field of audio processing. The method includes: extracting audio features from an audio file to generate a number of audio feature sequences; performing feature difference analysis on the several audio feature sequences respectively to output a number of difference feature sets; performing slice length retrieval analysis based on the several difference feature sets to output an adapted slice length; and performing audio slicing processing on the audio file according to the adapted slice length. The method aims to solve the technical problem that existing audio slicing processing methods have difficulty in accurately identifying the true boundaries of semantics or content structures, resulting in insufficient accuracy in audio slicing processing. By dynamically adjusting the slice length, semantic turning points and content change areas can be more accurately identified, structurally adaptive audio slicing processing can be achieved, and the recognition accuracy of slice boundaries and overall processing accuracy can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of audio processing, and in particular to an audio slicing processing method, system and storage medium for audio feature analysis. Background Art

[0002] With the development of application scenarios such as speech recognition, audio retrieval, intelligent customer service, and multimedia content analysis, higher requirements are placed on the automated and structured processing of audio data. Among them, audio slicing is a key link in audio preprocessing, and its slicing quality directly affects the effectiveness of subsequent tasks such as semantic understanding, sentiment analysis, and feature modeling.

[0003] Existing audio slicing methods are mostly based on simple rules such as fixed time length division or silence detection. Although basically usable slicing effects can be obtained in some homogeneous speech signals, it is often difficult to accurately identify the true content boundaries when processing audio with complex semantics, drastic intonation changes, or changing background sounds. Especially in cases of semantic turning points, speaker changes, music and speech mixing, existing methods are unable to adaptively adjust the slicing strategy according to the actual changes in audio features, resulting in large deviations in slicing boundaries and damaged semantic integrity, which in turn affects the recognition accuracy and processing efficiency of subsequent models.

[0004] Therefore, there is an urgent need for an audio slicing processing method that can perform intelligent analysis based on the changing rules of audio features and realize dynamic slice length adjustment, so as to improve the structural accuracy and processing precision of audio content slicing and meet the high-quality intelligent analysis needs in complex audio scenarios. Summary of the Invention

[0005] The present invention aims to provide an audio slice processing method, system, and storage medium for audio feature analysis, which are intended to address the technical problem that existing audio slice processing methods have difficulty accurately identifying the true boundaries of semantics or content structure, resulting in insufficient accuracy in audio slice processing. The present invention includes:

[0006] In a first aspect, the present invention provides an audio slicing processing method for audio feature analysis, comprising: extracting audio features from an audio file to generate a plurality of audio feature sequences; performing feature difference analysis on the plurality of audio feature sequences respectively to output a plurality of difference feature sets; performing slice length retrieval analysis based on the plurality of difference feature sets to output an adapted slice length; and performing audio slicing processing on the audio file according to the adapted slice length.

[0007] Preferably, the audio slicing processing method for audio feature analysis also includes: configuring audio feature indicators, wherein the audio feature indicators include at least short-time energy, zero crossing rate, MFCC mean and first-order difference, spectrum centroid, spectrum contrast and pitch; according to the audio feature indicators, audio features of the audio file are extracted to obtain several audio feature sequences.

[0008] Preferably, the audio slicing processing method for audio feature analysis also includes: randomly selecting any audio feature sequence from the several audio feature sequences as the audio feature sequence to be analyzed; performing feature difference analysis on adjacent audio features in the audio feature sequence to be analyzed, outputting a first difference feature set, and analyzing in sequence to obtain several difference feature sets.

[0009] Preferably, the audio slicing processing method for audio feature analysis also includes: selecting the first feature in the audio feature sequence to be analyzed as the first key feature, and selecting the adjacent feature of the first key feature as the second feature; performing a similarity comparison on the first key feature and the second feature to obtain a first similarity, and subtracting the first similarity from 1 to obtain a first difference; if the first difference is less than or equal to a preset difference threshold, discarding the second feature, and performing iterative feature comparison based on the first key feature in the feature order; if the first difference is greater than the preset difference threshold, setting the second feature as the second key feature, and performing iterative feature comparison based on the second key feature in the feature order until the feature comparison is completed, and obtaining multiple key features and multiple feature differences; performing feature difference analysis based on the multiple key features and multiple feature differences, and outputting a first difference feature set.

[0010] Preferably, the audio slicing processing method for audio feature analysis also includes: counting the feature quantity of the multiple key features to obtain a first difference feature quantity; performing mean calculation on the multiple feature differences to output a first difference mean; performing feature fluctuation analysis on the multiple feature differences to output a first difference fluctuation, wherein the difference fluctuation is the ratio of the difference standard deviation to the first difference mean; summarizing the first difference feature quantity, the first difference mean and the first difference fluctuation to obtain a first difference feature set.

[0011] Preferably, the audio slice processing method for audio feature analysis also includes: randomly selecting any difference feature set from the several difference feature sets as the first difference feature set; configuring a feature tolerance threshold, expanding the first difference feature set, and obtaining a first difference feature interval set; using the first difference feature interval set as a constraint condition, using audio slice processing as a guide word, using big data to perform slice length retrieval and analysis, and outputting an adapted slice length.

[0012] Preferably, the audio slice processing method for audio feature analysis also includes: taking the first difference feature interval set as a constraint condition, taking audio slice processing as a guide word, using big data to perform slice length retrieval analysis, and outputting a first slice length set, wherein the first slice length is marked with a historical slice accuracy rate; screening the first slice length whose historical slice accuracy rate is greater than a preset slice accuracy rate threshold and setting it as a standard slice length to obtain a standard slice length set, and selecting the majority of the standard slice length set as an adaptive slice length.

[0013] In the second aspect, the present invention also provides an audio slice processing system for audio feature analysis, which is used to execute an audio slice processing method for audio feature analysis as described in the first aspect, including: an audio feature extraction module, used to extract audio features of an audio file and generate a number of audio feature sequences; a feature difference analysis module, used to perform feature difference analysis on the several audio feature sequences respectively, and output a number of difference feature sets; a slice length retrieval module, used to perform slice length retrieval analysis based on the several difference feature sets, and output an adapted slice length; an audio slice processing module, used to perform audio slicing processing on the audio file according to the adapted slice length.

[0014] In a third aspect, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed, the steps of the method described in any one of the first aspects are implemented.

[0015] The embodiments of the present invention include the following advantages:

[0016] The method extracts audio features from audio files to generate several audio feature sequences. Feature difference analysis is then performed on each of these audio feature sequences, outputting several difference feature sets. Slice length retrieval and analysis is then performed based on these difference feature sets, outputting an adapted slice length. Finally, the audio file is sliced ​​according to the adapted slice length. In other words, by extracting multidimensional audio features and performing difference analysis, dynamically adjusting the slice length, it is possible to more accurately identify semantic turning points and areas of content change, achieving structurally adaptive audio slicing, and effectively improving the accuracy of slice boundary recognition and overall processing precision. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A flowchart of the steps of an audio slicing processing method for audio feature analysis according to the present invention;

[0018] Figure 2 This is a structural diagram of an audio slice processing system for audio feature analysis according to the present invention;

[0019] Figure 3 A schematic structural diagram of a computer-readable storage medium provided by the present invention.

[0020] Description of reference numerals:

[0021] Audio feature extraction module 11, feature difference analysis module 12, slice length retrieval module 13, audio slice processing module 14, computer readable storage medium 600, computer program 611. DETAILED DESCRIPTION

[0022] This invention addresses the technical problem of existing audio slicing methods struggling to accurately identify the true boundaries of semantics or content structure, resulting in insufficient audio slicing accuracy, by providing an audio feature analysis audio slicing processing method, system, and storage medium. By extracting multidimensional audio features and performing differential analysis, and dynamically adjusting the slice length, it is possible to more accurately identify semantic turning points and areas of content change, achieving structurally adaptive audio slicing processing, effectively improving both the accuracy of slice boundary recognition and overall processing precision.

[0023] Below, the technical solutions of the present invention will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention, rather than all, are shown in the accompanying drawings.

[0024] For example 1, please refer to the attached Figure 1 The present invention provides an audio slice processing method for audio feature analysis, which is applied to an audio slice processing system for audio feature analysis and specifically includes the following steps:

[0025] S10: extracting audio features from the audio file to generate several audio feature sequences.

[0026] Furthermore, step S10 of the present invention further includes:

[0027] S11: Configuring audio feature indicators, wherein the audio feature indicators include at least short-time energy, zero crossing rate, MFCC mean and first-order difference, spectrum centroid, spectrum contrast and pitch; S12: Extracting audio features from the audio file according to the audio feature indicators to obtain several audio feature sequences.

[0028] Specifically, first, before performing audio slicing processing, it is necessary to first set a number of audio feature indicators with discrimination and structural sensitivity for subsequent extraction and analysis of audio content changes. The audio feature indicators include at least short-time energy, zero crossing rate, MFCC mean and first-order difference, spectral centroid, spectral contrast and pitch. Among them, short-time energy characterizes the energy intensity of the audio signal in a short time window, which can be used to reflect the loudness change of the speech signal, and judge the boundary information such as the presence or absence of pronunciation, the beginning and end of the segment; the zero crossing rate indicates the number of times the audio signal crosses the zero point per unit time, which is suitable for analyzing the conversion between speech and silence, unvoiced and voiced sounds, and helps to detect pauses and intonation fluctuations. The MFCC mean provides stable features such as the overall timbre and resonance peak structure of the audio signal; the first-order difference reveals the dynamic change trend of these features over time, which is conducive to detecting speech turning points or semantic boundaries; the spectral centroid describes the "center of gravity" position of the spectrum, reflecting the brightness and timbre structure of the sound, and is suitable for judging the switching boundaries between speech segments and non-speech segments, and different sound sources; the spectral contrast measures the amplitude difference between different frequency bands, which can distinguish complex speech from background noise or music content, and enhance the robustness of boundary recognition; the pitch represents the change of the main frequency in the audio, which can be used to analyze semantically significant areas such as the speaker's tone, emotional fluctuations, and sentence end symbols, which helps to determine the cut point. Through the construction of features in the above multiple dimensions, it can comprehensively reflect the energy structure, spectral morphology, speech intonation, semantic changes and other characteristics of the audio, providing support for subsequent difference analysis and dynamic slicing decisions.

[0029] Then, after configuring the feature indicators, the input audio file is divided into frames (usually 10 to 30 milliseconds per frame, with overlap), and the above-mentioned feature indicators are calculated for each frame to obtain the corresponding time series data, where each audio feature indicator corresponds to a feature sequence; each item in each feature sequence corresponds to the feature value of the audio frame at that time point; all feature sequences together constitute a multi-dimensional feature matrix, reflecting the dynamic evolution of audio over time; for example: short-time energy sequences can show obvious differences between speech segments and pause segments; spectral centroid sequences can identify the location of timbre changes; these feature sequences will serve as the input data source for subsequent difference feature extraction and adaptive slicing decisions.

[0030] S20: performing feature difference analysis on the plurality of audio feature sequences respectively, and outputting a plurality of difference feature sets.

[0031] Furthermore, step S20 of the present invention further includes:

[0032] S21: Randomly select any audio feature sequence from the plurality of audio feature sequences as the audio feature sequence to be analyzed.

[0033] Specifically, first, any audio feature sequence is randomly selected from the plurality of audio feature sequences and set as the audio feature sequence to be analyzed.

[0034] S22: performing feature difference analysis on adjacent audio features in the audio feature sequence to be analyzed, outputting a first difference feature set, and sequentially analyzing to obtain a plurality of difference feature sets.

[0035] Furthermore, step S22 of the present invention further includes:

[0036] S221: Select the first feature in the audio feature sequence to be analyzed as the first key feature, and select the adjacent feature of the first key feature as the second feature; S222: Perform a similarity comparison on the first key feature and the second feature to obtain a first similarity, and subtract the first similarity from 1 to obtain a first difference; S223: If the first difference is less than or equal to a preset difference threshold, discard the second feature, and use the first key feature as a reference to perform iterative feature comparison in feature order; S224: If the first difference is greater than the preset difference threshold, set the second feature as the second key feature, and use the second key feature as a reference to perform iterative feature comparison in feature order until the feature comparison is completed, and obtain multiple key features and multiple feature differences.

[0037] Specifically, in order to analyze the degree of change between frames in an audio feature sequence, the feature value of the current frame in the audio feature sequence to be analyzed is first selected as the first key feature, and the feature value of one or more adjacent frames is set as the second feature to form a local feature pair in continuous time. Subsequently, by calculating the similarity between the two features, for example, using cosine similarity, Euclidean distance normalization results or other similarity measurement methods, a quantified first similarity value is obtained. In order to reflect the degree of "change" between features rather than the degree of similarity, the "1 minus similarity" method is further adopted to convert it into a first difference, that is, the intensity of the feature change at that position. This difference can be used as an important basis for subsequent judgment of the boundaries of audio slices. The higher the difference, the more likely there is a semantic transition or structural change at that position.

[0038] Next, the algorithm determines whether the calculated first difference is less than or equal to a preset difference threshold, which measures whether adjacent features have changed significantly. If the difference is small, the feature changes between the current frame and the adjacent frame are not significant, potentially belonging to the same semantic or acoustic segment. Therefore, the current second feature can be discarded, meaning it fails to form a new valid slice boundary. The algorithm then iterates backward, using the first key feature as a benchmark. The next adjacent feature is selected as the new second feature, and similarity comparison and difference calculation are repeated. This process continues iteratively. If the first difference is greater than the preset difference threshold, it indicates a significant change between the features of the two adjacent frames, possibly corresponding to a structural boundary such as a semantic transition, a speech segment switch, or a sound source shift. Therefore, the second feature is set as the new key feature (the second key feature) and used as the new comparison benchmark. Iterative feature comparisons are then performed with subsequent adjacent frames in chronological order. Each comparison generates a similarity and a corresponding difference. This approach continuously tracks and records multiple key feature points and their corresponding feature differences across the entire audio feature sequence, forming a set of difference analysis results that reflect the changing trends of the audio content, providing fundamental support for subsequent adaptive slice length decisions and boundary location.

[0039] S225: Perform feature difference analysis based on the multiple key features and the multiple feature difference degrees, and output a first difference feature set.

[0040] Furthermore, step S225 of the present invention further includes:

[0041] S2251: Count the number of features of the multiple key features to obtain a first number of difference features; S2252: Calculate the mean of the multiple feature differences and output a first difference mean; S2253: Perform feature fluctuation analysis on the multiple feature differences and output a first difference fluctuation, wherein the difference fluctuation is the ratio of the difference standard deviation to the first difference mean; S2254: Summarize the first number of difference features, the first difference mean and the first difference fluctuation to obtain a first difference feature set.

[0042] Specifically, first, the total number of these key features is counted to obtain the first difference feature number reflecting the number of change points; then, the mean of all feature difference values ​​is calculated to obtain the first difference mean of the overall change degree; then, the fluctuation of these differences is analyzed, and the first difference fluctuation reflecting the stability and volatility of the change is obtained by calculating the ratio of the standard deviation of the difference to the mean; finally, these three indicators - the number of difference features, the mean difference and the difference fluctuation - are summarized and integrated to form the first difference feature set for subsequent analysis, so as to comprehensively reflect the change rules and structural characteristics in the audio feature sequence, and provide a quantitative basis for the decision of the dynamic slice length.

[0043] S30: Performing slice length retrieval analysis based on the plurality of difference feature sets, and outputting an adapted slice length.

[0044] Furthermore, step S30 of the present invention further includes:

[0045] S31: randomly selecting any one of the plurality of difference feature sets as a first difference feature set; S32: configuring a feature tolerance threshold, expanding the first difference feature set, and obtaining a first difference feature interval set.

[0046] Specifically, first, one is randomly selected from multiple difference feature sets as the first difference feature set for the current analysis to ensure the diversity and robustness of the processing. Subsequently, based on the pre-configured feature tolerance threshold, the difference feature set is expanded and adjusted in scope to form a first difference feature interval set containing possible change boundaries, wherein the feature tolerance threshold refers to the threshold limit that allows feature changes within a certain range to be regarded as similar or continuous when analyzing audio feature differences. It is used to tolerate small fluctuations in audio features in a short time or between adjacent frames, and avoid excessive cutting or misjudgment of boundaries due to subtle changes. By setting an appropriate tolerance threshold, the recognition interval of the difference feature can be expanded, making slice boundary detection more robust and flexible, thereby improving the accuracy and robustness of audio slicing processing, and providing a more flexible and accurate basis for subsequent dynamic slice length decisions.

[0047] S33: Using the first difference feature interval set as a constraint condition and audio slice processing as a guide word, use big data to perform slice length retrieval and analysis, and output the adapted slice length.

[0048] Furthermore, step S33 of the present invention further includes:

[0049] S331: Taking the first difference feature interval set as the constraint condition and audio slicing processing as the guide word, big data is used to perform slice length retrieval and analysis, and a first slice length set is output, wherein the first slice length is identified with a historical slice accuracy rate; S332: The first slice length whose historical slice accuracy rate is greater than the preset slice accuracy rate threshold is selected and set as the standard slice length, a standard slice length set is obtained, and the majority of the standard slice length set is selected and set as the adaptive slice length.

[0050] Specifically, using the previously obtained first difference feature interval set as a constraint condition, combined with the specific task guidance of audio slicing processing, big data technology is used to retrieve and analyze a large number of historical audio slicing cases; by matching these difference feature intervals with the corresponding slice lengths and their slicing effects in historical data, the system generates a set of multiple candidate slice lengths, called the first slice length set, in which each slice length not only represents a slice length, but also comes with the actual slicing accuracy achieved by this length in the historical slicing process, reflecting the quality of its slicing effect.

[0051] Subsequently, based on a preset slicing accuracy threshold, the lengths in the first slicing length set whose slicing accuracy exceeds the threshold are screened to form a standard slicing length set. This screening process ensures that only slice lengths with good performance and stable slicing effects are retained. Finally, by counting the frequency of occurrence of slice lengths in the standard slicing length set, the slice length with the highest frequency is selected as the final adaptive slice length. This ensures that the selected slice length not only conforms to historical performance but also has strong representativeness and applicability, thereby achieving dynamic, adaptive, and high-precision slicing processing of the current audio file.

[0052] S40: Perform audio slicing processing on the audio file according to the adapted slicing length.

[0053] Specifically, the audio file is divided into multiple continuous segments using the adaptive slice length as the time window, with the duration of each segment roughly meeting the length standard. This dynamically adjusted slicing method can more accurately correspond to semantic or structural changes in the audio content, avoiding the boundary misalignment or content fragmentation problems caused by traditional fixed-length slicing, thereby improving the overall quality of the audio slices and the effectiveness of subsequent processing.

[0054] In summary, the audio slicing processing method for audio feature analysis provided by the present invention has the following technical effects:

[0055] The method extracts audio features from audio files to generate several audio feature sequences. Feature difference analysis is then performed on each of these audio feature sequences, outputting several difference feature sets. Slice length retrieval and analysis is then performed based on these difference feature sets, outputting an adapted slice length. Finally, the audio file is sliced ​​according to the adapted slice length. In other words, by extracting multidimensional audio features and performing difference analysis, dynamically adjusting the slice length, it is possible to more accurately identify semantic turning points and areas of content change, achieving structurally adaptive audio slicing, and effectively improving the accuracy of slice boundary recognition and overall processing precision.

[0056] In the second embodiment, based on the same inventive concept as the audio slice processing method for audio feature analysis in the above embodiment, the present invention also provides an audio slice processing system for audio feature analysis, see the attached Figure 2 , including: an audio feature extraction module 11, used to extract audio features from audio files and generate several audio feature sequences; a feature difference analysis module 12, used to perform feature difference analysis on the several audio feature sequences respectively and output several difference feature sets; a slice length retrieval module 13, used to perform slice length retrieval analysis based on the several difference feature sets and output an adapted slice length; an audio slice processing module 14, used to perform audio slice processing on the audio file according to the adapted slice length.

[0057] Furthermore, the audio slicing processing system for audio feature analysis is also used to: configure audio feature indicators, wherein the audio feature indicators include at least short-time energy, zero crossing rate, MFCC mean and first-order difference, spectrum centroid, spectrum contrast and pitch; according to the audio feature indicators, perform audio feature extraction on the audio file to obtain several audio feature sequences.

[0058] Furthermore, the audio slicing processing system for audio feature analysis is also used to: randomly select any audio feature sequence from the several audio feature sequences as the audio feature sequence to be analyzed; perform feature difference analysis on adjacent audio features in the audio feature sequence to be analyzed, output a first difference feature set, and analyze in sequence to obtain several difference feature sets.

[0059] Furthermore, the audio slicing processing system for audio feature analysis is also used to: select the first feature in the audio feature sequence to be analyzed as the first key feature, and select the adjacent feature of the first key feature as the second feature; perform a similarity comparison on the first key feature and the second feature to obtain a first similarity, and subtract the first similarity from 1 to obtain a first difference; if the first difference is less than or equal to a preset difference threshold, discard the second feature, and use the first key feature as a reference to perform iterative feature comparison in feature order; if the first difference is greater than the preset difference threshold, set the second feature as the second key feature, and use the second key feature as a reference to perform iterative feature comparison in feature order until the feature comparison is completed, and obtain multiple key features and multiple feature differences; perform feature difference analysis based on the multiple key features and multiple feature differences, and output a first difference feature set.

[0060] Furthermore, the audio slicing processing system for audio feature analysis is also used to: count the feature quantity of the multiple key features to obtain a first difference feature quantity; perform mean calculation on the multiple feature differences to output a first difference mean; perform feature fluctuation analysis on the multiple feature differences to output a first difference fluctuation, wherein the difference fluctuation is the ratio of the difference standard deviation to the first difference mean; summarize the first difference feature quantity, the first difference mean and the first difference fluctuation to obtain a first difference feature set.

[0061] Furthermore, the audio slice processing system for audio feature analysis is also used to: randomly select any difference feature set from the several difference feature sets and set it as the first difference feature set; configure a feature tolerance threshold, expand the first difference feature set, and obtain a first difference feature interval set; use the first difference feature interval set as a constraint condition, use audio slice processing as a guide word, use big data to perform slice length retrieval and analysis, and output an adapted slice length.

[0062] Furthermore, the audio slice processing system for audio feature analysis is also used to: use the first difference feature interval set as a constraint condition, audio slice processing as a guide word, use big data to perform slice length retrieval analysis, and output a first slice length set, wherein the first slice length is identified by a historical slice accuracy rate; filter the first slice length whose historical slice accuracy rate is greater than a preset slice accuracy rate threshold and set it as a standard slice length to obtain a standard slice length set, and select the majority of the standard slice length set as an adaptive slice length.

[0063] For example three, please refer to Figure 3 , Figure 3 Schematic diagram of an embodiment of a computer-readable storage medium provided in an embodiment of the present invention. Figure 3 As shown, this embodiment provides a computer-readable storage medium 600, on which a computer program 611 is stored. When the computer program 611 is executed by a processor, the following steps are implemented: audio feature extraction is performed on an audio file to generate a plurality of audio feature sequences; feature difference analysis is performed on the plurality of audio feature sequences respectively, and a plurality of difference feature sets are output; slice length retrieval analysis is performed based on the plurality of difference feature sets, and an adapted slice length is output; and audio slicing processing is performed on the audio file according to the adapted slice length.

[0064] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0065] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0066] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0067] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0068] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0069] Although preferred embodiments of the present invention have been described, additional changes and modifications to these embodiments may occur to those skilled in the art once the basic inventive concepts become known.

[0070] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the present invention and its equivalents, the present invention is intended to include these modifications and variations.

Claims

1. An audio slice processing method for audio feature analysis, characterized in that the method include: Extract audio features from audio files to generate several audio feature sequences; Performing feature difference analysis on the plurality of audio feature sequences respectively, and outputting a plurality of difference feature sets; Performing slice length retrieval analysis based on the plurality of difference feature sets, and outputting an adapted slice length; Performing audio slicing processing on the audio file according to the adapted slice length; Among them, audio features are extracted from the audio file to generate several audio feature sequences, including: Configuring audio feature indicators, wherein the audio feature indicators include at least short-time energy, zero crossing rate, MFCC mean and first-order difference, spectrum centroid, spectrum contrast and pitch; Extracting audio features from the audio file according to the audio feature indicators to obtain a plurality of audio feature sequences; The slice length retrieval analysis is performed based on the plurality of difference feature sets, and the adapted slice length is output, including: Randomly selecting any one of the plurality of difference feature sets as a first difference feature set; Configuring a feature tolerance threshold, and expanding the first difference feature set to obtain a first difference feature interval set; Using the first difference feature interval set as a constraint condition and audio slicing processing as a guide word, using big data to perform slice length retrieval analysis and output an adapted slice length; The method uses the first difference feature interval set as a constraint condition, audio slicing processing as a guide word, and big data to perform slice length retrieval and analysis, and output an adapted slice length, including: Using the first difference feature interval set as a constraint condition and audio slicing processing as a guide word, using big data to perform slice length retrieval analysis, and outputting a first slice length set, wherein the first slice length is identified by a historical slice accuracy rate; The first slice length whose historical slice accuracy is greater than a preset slice accuracy threshold is screened and set as the standard slice length to obtain a standard slice length set, and the mode in the standard slice length set is selected and set as the adaptive slice length.

2. The audio slicing processing method for audio feature analysis according to claim 1, characterized in that: Perform feature difference analysis on the plurality of audio feature sequences respectively, and output a plurality of difference feature sets, including: Randomly selecting any audio feature sequence from the plurality of audio feature sequences as the audio feature sequence to be analyzed; Feature difference analysis is performed on adjacent audio features in the audio feature sequence to be analyzed, and a first difference feature set is output, and a plurality of difference feature sets are obtained by sequential analysis.

3. The audio slicing processing method for audio feature analysis according to claim 2, characterized in that: Performing feature difference analysis on adjacent audio features in the audio feature sequence to be analyzed, and outputting a first difference feature set, including: Selecting a first feature in the audio feature sequence to be analyzed as a first key feature, and selecting an adjacent feature of the first key feature as a second feature; Performing a similarity comparison on the first key feature and the second feature to obtain a first similarity, and subtracting the first similarity from 1 to obtain a first difference; If the first difference is less than or equal to a preset difference threshold, the second feature is discarded, and the first key feature is used as a benchmark, and feature comparison is performed iteratively in feature order; If the first difference is greater than the preset difference threshold, the second feature is set as the second key feature, and the second key feature is used as a benchmark to perform iterative feature comparison in feature order until the feature comparison is completed, thereby obtaining multiple key features and multiple feature difference degrees; Perform feature difference analysis based on the multiple key features and the multiple feature difference degrees, and output a first difference feature set.

4. The audio slicing processing method for audio feature analysis according to claim 3, characterized in that: Performing feature difference analysis based on the multiple key features and the multiple feature difference degrees, and outputting a first difference feature set, including: Counting the number of features of the plurality of key features to obtain a first number of difference features; Calculating the mean of the plurality of feature differences and outputting a first mean of the differences; Performing feature fluctuation analysis on the multiple feature differences and outputting a first difference fluctuation, wherein the difference fluctuation is a ratio of a difference standard deviation to a first difference mean; The first difference feature quantity, the first difference degree mean, and the first difference fluctuation are summarized to obtain a first difference feature set.

5. An audio slice processing system for audio feature analysis, characterized in that: The steps for implementing the audio slice processing method for audio feature analysis according to any one of claims 1 to 4 include: An audio feature extraction module is used to extract audio features from audio files and generate several audio feature sequences; A feature difference analysis module, configured to perform feature difference analysis on the plurality of audio feature sequences respectively and output a plurality of difference feature sets; a slice length retrieval module, configured to perform slice length retrieval analysis based on the plurality of difference feature sets and output an adapted slice length; The audio slicing processing module is used to perform audio slicing processing on the audio file according to the adapted slicing length.

6. A non-transitory computer-readable storage medium, characterized in that The storage medium stores a computer software program, which, when executed by a processor, implements the steps of an audio slice processing method for audio feature analysis as described in any one of claims 1 to 4.

Images