A sound data processing method and system
The denoising of sound data through empirical modal decomposition and clustering technology solves the problem that bandpass filtering cannot completely separate noise and signal, achieves better noise removal effect, and improves the quality and clarity of sound data.
Patent Information
- Application Number
- CN202411054562.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-02
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-08-02
AI Technical Summary
The existing bandpass filtering technology cannot completely separate the spectrum characteristics of noise and signal, resulting in partially deleting or distorting of sound data during processing, affecting the quality and clarity of sound data.
The empirical modal decomposition (EMD) algorithm is used to decompose the sound data into multiple IMF components and residual components. The noise influence degree of local maximum and minimum points is determined through clustering, and the noise influence degree of target extreme points is corrected. , and filter the IMF components according to the size of the filter window, and finally obtain the denoised sound data through EMD reconstruction.
Effectively remove noise, especially noise characteristics at different frequencies, more comprehensive consideration is given, improving the noise removal effect of sound data and ensuring the quality and clarity of the signal.
Smart Images

Figure CN118609586B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of sound data processing, and in particular to a sound data processing method and system. Background Art
[0002] Sound is the sound wave generated by the vibration of an object. Noise will be mixed into the sound signal, reducing the clarity and accuracy of the sound signal and interfering with the extraction and analysis of sound features. Therefore, it is necessary to denoise the sound data.
[0003] Currently, sound data is usually denoised through bandpass filtering. Bandpass filtering can remove noise within a specific frequency range, but noise has energy distribution in each frequency band. However, bandpass filtering may not be able to completely separate the spectral characteristics of noise and signal, resulting in partial deletion or distortion of the signal during the processing process, thereby affecting the quality and clarity of the sound data. Summary of the invention
[0004] In order to solve the above problems, the present invention provides a sound data processing method and system.
[0005] In a first aspect, the present invention provides a method for processing sound data, which adopts the following technical solution:
[0006] A method for processing sound data, comprising the steps of:
[0007] Collecting sound data; decomposing the sound data into multiple IMF components and a residual component;
[0008] Determining the noise influence degree of each local maximum point in the IMF component, including: clustering all local maximum points in the IMF component, and determining the noise influence degree of the local maximum point according to the difference between the local maximum point and the remaining local maximum points in the category to which the local maximum point belongs, and the number of local maximum points in the category to which the local maximum point belongs;
[0009] Determine the degree of noise influence of each local minimum point in the IMF component;
[0010] The noise influence degree of the target extreme point of the IMF component is corrected according to the noise influence degree of the local extreme point corresponding to the target extreme point in each of the other IMF components to obtain the final noise influence degree of the target extreme point; the filtering window size of each data point in the IMF component is obtained according to the final noise influence degree of each local extreme point in the IMF component; and the IMF component is filtered according to the filtering window size;
[0011] The filtering results of all IMF components and the residual components are reconstructed by EMD to obtain the denoised sound data.
[0012] Preferably, the method for obtaining the noise influence degree of the local maximum point is:
[0013] The category obtained by clustering all local maximum points in the IMF component is recorded as the first category; any local maximum point in the IMF component is taken as the target maximum point, and the noise influence degree of the target maximum point satisfies the expression:
[0014] ;
[0015] in, Indicates the degree of noise influence on the target maximum point; Indicates the number of local maximum points contained in the first category to which the target maximum point belongs; Represents the mean of the number of local maximum points contained in all first categories; Indicates the coverage size of the first category to which the target maximum point belongs; Indicates the amplitude of the target maximum point; Indicates the category to which the target maximum point belongs. The amplitude of the local maximum point; represents the natural exponential function;
[0016] The method for obtaining the coverage range size is: the moments corresponding to the local maximum point with the smallest number of moments and the maximum point with the largest number of moments in the first category to which the target maximum point belongs constitute a time interval, and the number of local maximum points in the IMF component located in the time interval is used as the coverage range size of the first category to which the target maximum point belongs.
[0017] Preferably, the step of correcting the noise influence degree of the target extreme point according to the noise influence degree of the local extreme point corresponding to the target extreme point of the IMF component in each of the other IMF components to obtain the final noise influence degree of the target extreme point includes:
[0018] The noise influence degree of the target extreme point of the target IMF component is corrected according to the noise influence degree of the local extreme point corresponding to the target extreme point of the target IMF component in each of the other IMF components, and the corrected noise influence degree of the target extreme point of the target IMF component under the other IMF components is obtained; the weighted summation of the corrected noise influence degrees of the target extreme point of the target IMF component under the other IMF components is performed to obtain the final noise influence degree of the target extreme point of the target IMF component.
[0019] Preferably, the degree of influence of the corrected noise of the target extreme point of the target IMF component under the remaining IMF components satisfies the expression:
[0020] ;
[0021] The target IMF component number is represented by To express, Indicates that the target extreme value point of the target IMF component is The influence degree of the corrected noise under each IMF component, is the sequence number of the remaining IMF components, and , represents the number of IMF components; Indicates the degree of noise influence of the target extreme point of the target IMF component; Indicates that the target extreme value point of the target IMF component is The degree of noise influence of the corresponding local extreme point in each IMF component; represents the natural exponential function.
[0022] Preferably, the weighted summing of the corrected noise influence degrees of the target extreme value point of the target IMF component under the remaining IMF components to obtain the final noise influence degree of the target extreme value point of the target IMF component also includes:
[0023] Will As the weight of each IMF component, Indicates the number of IMF components.
[0024] Preferably, the weighted summing of the corrected noise influence degrees of the target extreme value point of the target IMF component under the remaining IMF components to obtain the final noise influence degree of the target extreme value point of the target IMF component also includes:
[0025] The weights of each IMF component satisfy the expression:
[0026] ;
[0027] in, Indicates The weights of the IMF components; Indicates the serial number of the target IMF component; is the sequence number of the remaining IMF components, and , represents the number of IMF components; represents the time corresponding to the target extreme point of the target IMF component; represents the time corresponding to the local extreme point of the target extreme point of the target IMF component in the kth IMF component; is an S-type function; represents the natural exponential function.
[0028] Preferably, the filter window size of each data point in the IMF component satisfies the expression:
[0029] ;
[0030] in, Indicates the first component of the target IMF The filter window size for each data point; Indicates the serial number of the target IMF component; Indicates the target IMF component to The final noise influence of the local extreme point with the smallest distance to the data points; represents a hyperparameter; Represents the ceiling function.
[0031] In a second aspect, the present invention provides a sound data processing system, which adopts the following technical solution:
[0032] A sound data processing system comprises: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned sound data processing method is implemented.
[0033] By adopting the above technical solution, the above-mentioned sound data processing method is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a terminal device based on the memory and the processor for easy use.
[0034] The present invention has the following technical effects:
[0035] Noise has energy distribution in each frequency band. The present invention decomposes the sound data into multiple IMF components, filters each component separately, takes into account the characteristics of noise at different frequencies, and has a better denoising effect. The present invention corrects the noise influence degree of the target extreme point according to the noise influence degree of the local extreme point corresponding to the target extreme point of the IMF component in each of the other IMF components, so that the final noise influence degree of the target extreme point is more accurate, further improving the subsequent denoising effect on the sound data. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts.
[0037] Figure 1 It is a flow chart of a method for processing sound data in an embodiment of the present invention. DETAILED DESCRIPTION
[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0039] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.
[0040] The embodiment of the present invention discloses a method for processing sound data. Figure 1 , comprising steps S1 to S4:
[0041] S1: Collect sound data.
[0042] An acoustic wave sensor is set near the target sound source and the sound data of the target sound source is collected by using the acoustic wave sensor.
[0043] At this point, the collection of sound data has been achieved.
[0044] S2: Decompose the sound data into multiple IMF components and a residual component, and determine the noise influence degree of each local extreme point in the IMF component.
[0045] Specifically, an Empirical Mode Decomposition (EMD) algorithm is used to decompose the sound data into a plurality of Intrinsic Mode Function (IMF) components and a residual component.
[0046] Take any IMF component as the target IMF component, obtain all local maximum points in the target IMF component, cluster all local maximum points using the density-based spatial clustering method (Density-Based Spatial Clustering of Applications with Noise, DBSCAN), and obtain multiple first categories. Count the number of local maximum points contained in each first category. Obtain the time at which the local maximum point with the smallest number of corresponding moments among all local maximum points contained in each first category is located as the minimum time of the first category. Obtain the time at which the local maximum point with the largest number of corresponding moments among all local maximum points contained in each first category is located as the maximum time of the first category. The minimum time and maximum time of each first category constitute the time interval of each first category. The time interval of the first category is a closed interval, the left boundary of the time interval is the minimum time of the first category, and the right boundary is the maximum time of the first category.
[0047] The number of local maximum points in the IMF component located in the time interval of each first category is taken as the coverage size of each first category.
[0048] Take any local maximum point as the target maximum point, and determine the noise impact of the target maximum point based on the difference between the target maximum point and the remaining local maximum points in the first category to which it belongs, and the number of local maximum points contained in the first category to which the target maximum point belongs:
[0049] ;
[0050] in, Indicates the degree of noise influence on the target maximum point; Indicates the number of local maximum points contained in the first category to which the target maximum point belongs; Represents the mean of the number of local maximum points contained in all first categories; Indicates the coverage size of the first category to which the target maximum point belongs; Indicates the amplitude of the target maximum point; Indicates the category to which the target maximum point belongs. The amplitude of the local maximum point; represents the natural exponential function.
[0051] The amplitude of normal sound varies between different voices or sound segments, but it shows a certain pattern and regularity as a whole, while noise data usually shows a more volatile, irregular or highly random waveform. Therefore, after the local maximum points in normal sound are clustered by DBSCAN, the local maximum points in each category are continuous in time series, while noise data will destroy the continuity in time series, making the local maximum points in the category discontinuous in time series. To reflect the continuity of the local maximum point in the first category to which the target maximum point belongs in time series, the greater the continuity of the local maximum point in the first category to which the target maximum point belongs, the greater the possibility that the local maximum point in the first category to which the target maximum point belongs is normal sound data. At this time, the greater the possibility that the target maximum point is normal sound data, the smaller the degree of noise influence of the target maximum point. If the continuity of the local maximum point in the first category to which the target maximum point belongs is greater, the probability that the local maximum point in the first category to which the target maximum point belongs is greater. The smaller it is, the greater the possibility that the first category to which the target maximum point belongs is affected by noise, and the greater the degree of noise influence on the target maximum point.
[0052] When the target maximum point is greatly affected by noise, the target maximum point does not conform to the fluctuation law of the surrounding local maximum points, resulting in the target maximum point being classified into a first category alone, or being classified into a first category with a small number of other local maximum points that are greatly affected by noise. Large, in order to avoid this situation The larger the value, the smaller the noise impact on the target maximum point. To make corrections, when the number of local maximum points contained in the first category to which the target maximum point belongs is smaller, the noise influence of the target maximum point is greater.
[0053] It reflects the difference in amplitude between the target maximum point and the other local maximum points in the first category to which the target maximum point belongs. When the difference is larger, the fluctuation law of the target maximum point and the other local maximum points in the first category to which it belongs differs more, the division effect of the first category to which the target maximum point belongs is worse, and the degree to which the target maximum point is affected by noise is greater. The less accurate the noise effect, the less accurate the result. right Make corrections, when the difference is larger, The smaller, the The greater the increase of , the greater the noise influence of the target maximum point; when the difference is smaller, the more similar the fluctuation law of the target maximum point and the other local maximum points in the first category to which it belongs is, the better the division effect of the first category to which the target maximum point belongs is. The more accurate the noise effect is, the more accurate the difference is. right When making corrections, the smaller the difference, The larger and closer to 1, the The smaller the increase of , the more the noise influence of the target maximum point depends on .
[0054] At this point, the noise influence degree of the target maximum point in the target IMF component is obtained.
[0055] Similarly, all local minimum points in the target IMF component are obtained, and DBSCAN clustering is performed on all local minimum points to obtain multiple second categories. The number of minimum points contained in each second category is counted to obtain the coverage size of each second category.
[0056] Take any local minimum point as the target minimum point, and determine the noise impact of the target minimum point based on the difference between the target minimum point and the remaining local minimum points in the second category to which it belongs, and the number of local minimum points contained in the second category to which the target minimum point belongs:
[0057] ;
[0058] in, Indicates the degree of noise influence on the target minimum point; Indicates the number of local minimum points contained in the second category to which the target minimum point belongs; represents the mean of the number of local minimum points contained in all second categories; The coverage size of the second category to which the target minimum point belongs; Indicates the amplitude of the target minimum point; Indicates the category to which the target minimum point belongs. The amplitude of the local minimum point; represents the natural exponential function.
[0059] Similarly, the noise influence degree of each local extreme point in each IMF component is obtained.
[0060] S3: Correct the noise influence degree of the target extreme point according to the noise influence degree of the local extreme point corresponding to the target extreme point of the IMF component in each of the other IMF components to obtain the final noise influence degree of the target extreme point.
[0061] It should be noted that each IMF component represents the oscillation behavior of a specific frequency and time scale in the sound data. Therefore, different IMF components reflect the changes in different frequencies in the sound data, and noise has energy distribution in each frequency band. Therefore, the present invention corrects the noise influence degree of the local extreme point in the target IMF component according to the noise influence degree of the local extreme point in different IMF components to obtain a more accurate noise assessment.
[0062] Specifically, all local extreme points in each IMF component are combined into a local extreme point sequence, and the local extreme point sequence of the target IMF component is matched with the local extreme point sequence of each other IMF component using the dynamic time warping (DTW) algorithm to obtain the local extreme point corresponding to each local extreme point of the target IMF component in each other IMF component.
[0063] Take any local extreme point in the target IMF component as the target extreme point, and correct the noise influence of the target extreme point of the target IMF component according to the noise influence of the local extreme point corresponding to the target extreme point of the target IMF component in each of the other IMF components, and obtain the corrected noise influence of the target extreme point of the target IMF component under the other IMF components:
[0064] ;
[0065] The target IMF component number is represented by To express, Indicates that the target extreme value point of the target IMF component is The influence degree of the corrected noise under each IMF component, is the sequence number of the remaining IMF components, and , represents the number of IMF components; Indicates the degree of noise influence of the target extreme point of the target IMF component; Indicates that the target extreme value point of the target IMF component is The degree of noise influence of the corresponding local extreme point in each IMF component; represents the natural exponential function.
[0066] when When it is less than the sequence number of the target IMF component, it means Compared with the target IMF component, the first IMF component reflects the higher-frequency change information in the sound data. Since the noise data has a more drastic, irregular or highly random waveform, the noise data is more likely to be more obvious in the higher-frequency IMF component and less obvious in the lower-frequency IMF component. Therefore, the noise impact of the noise data in the higher-frequency IMF component should be greater than the noise impact of the noise data in the lower-frequency IMF component. Therefore, when the target extreme point of the target IMF component is at When the noise influence degree of the corresponding local extreme point in the IMF component is less than the noise influence degree of the target extreme point of the target IMF component, the fluctuation of the target extreme point of the target IMF component is more likely to be the normal fluctuation of normal sound data. Correction is made so that the noise influence of the target extreme point of the target IMF component is reduced; when the target extreme point of the target IMF component is at When the noise influence degree of the corresponding local extreme point in an IMF component is greater than or equal to the noise influence degree of the target extreme point of the target IMF component, the noise influence degree of the target extreme point of the target IMF component is in line with expectations, and the noise influence degree of the target extreme point of the target IMF component is not corrected.
[0067] when When it is greater than the sequence number of the target IMF component, it means The IMF component reflects the lower-frequency change information in the sound data compared to the target IMF component. When the target extreme point of the target IMF component is at When the noise influence degree of the corresponding local extreme point in the first IMF component is less than the noise influence degree of the target extreme point of the target IMF component, the noise influence degree of the target extreme point of the target IMF component is in line with expectations. At this time, the noise influence degree of the target extreme point of the target IMF component is not corrected. When the target extreme point of the target IMF component is in the first When the noise influence degree of the corresponding local extreme point in the IMF component is greater than or equal to the noise influence degree of the target extreme point of the target IMF component, the noise influence degree of the target extreme point of the target IMF component is more likely to be smaller, so the use of The correction is performed so that the noise influence degree of the target extreme point of the target IMF component is increased.
[0068] The final noise influence degree of the target extreme point of the target IMF component is obtained according to the corrected noise influence degree of the target extreme point of the target IMF component under the other IMF components:
[0069] ;
[0070] in, Indicates the final noise influence degree of the target extreme point of the target IMF component; Indicates that the target extreme value point of the target IMF component is The influence degree of the corrected noise under each IMF component; Indicates The weights of the IMF components; Indicates the number of IMF components.
[0071] In one embodiment, Weights of IMF components The method to obtain is:
[0072] ;
[0073] in, Indicates The weights of the IMF components; represents the number of IMF components; As the The weights of the IMF components are set so that the weights of each IMF component other than the target IMF component are the same, that is, the reference degree of the final noise influence degree of each IMF component other than the target IMF component on the target extreme value point of the target IMF component is the same.
[0074] In another embodiment, Weights of IMF components The method to obtain is:
[0075] ;
[0076] in, Indicates The weights of the IMF components; Indicates the serial number of the target IMF component; is the sequence number of the remaining IMF components, and , represents the number of IMF components; represents the time corresponding to the target extreme point of the target IMF component; Indicates that the target extreme value point of the target IMF component is The time corresponding to the local extreme point in each IMF component; represents the natural exponential function. Used for Normalization.
[0077] Since the local extreme points corresponding to the target extreme points in each IMF component are obtained based on the DTW algorithm, the time points corresponding to the target extreme points and their corresponding local extreme points in each IMF component are not necessarily consistent. It reflects the difference between the target extreme point of the target IMF component and the target extreme point of the target IMF component in the The difference between the moments corresponding to the local extreme points in the IMF components. When the moments are closer, that is, the difference between the moments is smaller, the target extreme point is in the The more the noise influence of the local extreme point in the IMF component is, the more reference it has to the noise influence of the target extreme point. The larger the weight of the IMF component, the larger the difference between the moments. The less the noise influence of the local extreme point in the IMF component is on the target extreme point, the less reference it has. The smaller the weight of the IMF component.
[0078] Similarly, the noise influence degree of each local extreme point in each IMF component is corrected to obtain the final noise influence degree of each local extreme point in each IMF component.
[0079] S4: Obtain the filter window size of each data point in the IMF component according to the final noise influence degree of each local extreme point in the IMF component, filter the IMF component according to the filter window size, reconstruct the filter results of all IMF components and the residual components by EMD, and obtain the denoised sound data.
[0080] Specifically, the filtering window size of each data point in the target IMF component is obtained according to the final noise influence degree of each local extreme point in the target IMF component:
[0081] ;
[0082] in, Indicates the first component of the target IMF The filter window size for each data point; Indicates the serial number of the target IMF component; Indicates the target IMF component to The final noise influence of the local extreme point with the smallest distance to the data points; Represents a hyperparameter, which is used to adjust the basic size of the filter window. The empirical value is 2. Implementers can set the hyperparameter according to the actual implementation situation. The value of Represents the ceiling function.
[0083] Different IMF components reflect the changes of different frequencies in the sound data. The smaller the serial number of the IMF component, the higher the frequency of the IMF component. The high-frequency IMF component corresponds to the rapidly changing part of the sound data. The larger the serial number of the IMF component, the lower the frequency of the IMF component. The low-frequency IMF component corresponds to the slowly changing part of the sound data. The amplitude of normal sound varies between different voices or sound segments. Therefore, for high-frequency IMF components, in order to avoid the filtering interference of adjacent sound segments on the current sound segment, a smaller window is used for filtering. For low-frequency IMF components, a smaller window is used for filtering. Therefore, the present invention obtains the filtering window size based on the serial number of the IMF component.
[0084] Since local extreme points are affected by noise to different degrees, for local extreme points with a large degree of noise influence and the data points around them, a larger window should be used for filtering to remove the noise influence, and for local extreme points with a small degree of noise influence and the data points around them, a smaller window should be used for filtering to retain the characteristics of the data points themselves. As The index used to Make corrections when The larger the value, the larger the filter window. The smaller it is, the smaller the filter window is.
[0085] The target IMF component is filtered according to the filtering window size of each data point in the target IMF component to obtain the filtering result of the target IMF component. It should be noted that the filtering algorithm adopted by the present invention is mean filtering. In other implementations, the implementer can set the filtering algorithm according to the actual implementation situation.
[0086] Similarly, obtain the filtering results of all IMF components.
[0087] The filtering results of all IMF components and the residual components are reconstructed by EMD to obtain the denoised sound data.
[0088] At this point, the denoising of the sound data is achieved.
[0089] The embodiment of the present invention further discloses a sound data processing system, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a sound data processing method according to the present invention is implemented.
[0090] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.
[0091] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.
[0092] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for processing sound data, characterized in that: include: Collect sound data; Decompose the sound data into multiple IMF components and a residual component; Determining the noise influence degree of each local maximum point in the IMF component, including: clustering all local maximum points in the IMF component, and determining the noise influence degree of the local maximum point according to the difference between the local maximum point and the remaining local maximum points in the category to which the local maximum point belongs, and the number of local maximum points in the category to which the local maximum point belongs; Determine the degree of noise influence of each local minimum point in the IMF component; Any local extreme point in the IMF component is used as the target extreme point, and the local extreme point includes a local maximum point and a local minimum point; the noise influence degree of the target extreme point is corrected according to the noise influence degree of the local extreme point corresponding to the target extreme point of the IMF component in each of the other IMF components, so as to obtain the final noise influence degree of the target extreme point; the filtering window size of each data point in the IMF component is obtained according to the final noise influence degree of each local extreme point in the IMF component; and the IMF component is filtered according to the filtering window size; The filtering results of all IMF components and the residual components are reconstructed by EMD to obtain the denoised sound data.
2. A sound data processing method according to claim 1, characterized in that: The method for obtaining the noise influence degree of the local maximum point is: The category obtained by clustering all local maximum points in the IMF component is recorded as the first category; any local maximum point in the IMF component is taken as the target maximum point, and the noise influence degree of the target maximum point satisfies the expression: ; in, Indicates the degree of noise influence on the target maximum point; Indicates the number of local maximum points contained in the first category to which the target maximum point belongs; Represents the mean of the number of local maximum points contained in all first categories; Indicates the coverage size of the first category to which the target maximum point belongs; Indicates the amplitude of the target maximum point; Indicates the category to which the target maximum point belongs. The amplitude of the local maximum point; represents the natural exponential function; The method for obtaining the coverage range size is: the moments corresponding to the local maximum point with the smallest number of moments and the maximum point with the largest number of moments in the first category to which the target maximum point belongs constitute a time interval, and the number of local maximum points in the IMF component located in the time interval is used as the coverage range size of the first category to which the target maximum point belongs.
3. A sound data processing method according to claim 1, characterized in that: The step of correcting the noise influence degree of the target extreme point according to the noise influence degree of the local extreme point corresponding to the target extreme point of the IMF component in each of the other IMF components to obtain the final noise influence degree of the target extreme point includes: The noise influence degree of the target extreme point of the target IMF component is corrected according to the noise influence degree of the local extreme point corresponding to the target extreme point of the target IMF component in each of the other IMF components, and the corrected noise influence degree of the target extreme point of the target IMF component under the other IMF components is obtained; the weighted summation of the corrected noise influence degrees of the target extreme point of the target IMF component under the other IMF components is performed to obtain the final noise influence degree of the target extreme point of the target IMF component.
4. A sound data processing method according to claim 3, characterized in that: The influence degree of the corrected noise of the target extreme point of the target IMF component under the other IMF components satisfies the expression: ; The target IMF component number is represented by To express, Indicates that the target extreme value point of the target IMF component is The influence degree of the corrected noise under each IMF component, is the sequence number of the remaining IMF components, and , represents the number of IMF components; Indicates the degree of noise influence of the target extreme point of the target IMF component; Indicates that the target extreme value point of the target IMF component is The degree of noise influence of the corresponding local extreme point in each IMF component; represents the natural exponential function.
5. A sound data processing method according to claim 3 or 4, characterized in that: The weighted summation of the corrected noise influence degree of the target extreme value point of the target IMF component under the remaining IMF components to obtain the final noise influence degree of the target extreme value point of the target IMF component also includes: Will As the weight of each IMF component, Indicates the number of IMF components.
6. A sound data processing method according to claim 3 or 4, characterized in that: The weighted summation of the corrected noise influence degree of the target extreme value point of the target IMF component under the remaining IMF components to obtain the final noise influence degree of the target extreme value point of the target IMF component also includes: The weights of each IMF component satisfy the expression: ; in, Indicates The weights of the IMF components; Indicates the serial number of the target IMF component; is the sequence number of the remaining IMF components, and , represents the number of IMF components; represents the time corresponding to the target extreme point of the target IMF component; represents the time corresponding to the local extreme point of the target extreme point of the target IMF component in the kth IMF component; is an S-type function; represents the natural exponential function.
7. A sound data processing method according to claim 1, characterized in that: The filter window size of each data point in the IMF component satisfies the expression: ; in, Indicates the first component of the target IMF The filter window size for each data point; Indicates the serial number of the target IMF component; Indicates the target IMF component to The final noise influence of the local extreme point with the smallest distance to the data points; represents a hyperparameter; Represents the ceiling function.
8. A sound data processing system, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a sound data processing method according to any one of claims 1 to 7 is implemented.
Citation Information
Patent Citations
Support vector machine mechanical fault diagnosis method based on voting mechanism
CN108734192A
Audio data optimization and noise reduction method and system
CN117373471A
Intelligent noise reduction method for field bird sound data
CN117727313A
Adaptive learning noise reduction method
CN118335094A
Empirical mode decomposition for analyzing acoustical signals
US20030033094A1
