Audio data processing method and apparatus, chip, and electronic device

Abstract

An audio data processing method and apparatus (600), a chip, and an electronic device (800), relating to the technical field of data processing. The method comprises: obtaining a data matrix of audio data, wherein the data matrix is an MDFT data matrix corresponding to the audio data in an AVS encoder (101); determining a target parameter value of an exponentiation parameter corresponding to each column vector in the data matrix (102); on the basis of the target parameter value, performing power transformation on each column vector in the data matrix to obtain a power-transformed matrix corresponding to the data matrix, wherein the power transformation is used for concentrating the variance of the data matrix in the direction of a principal component, and increasing the feature percentage of the principal component in the data matrix (103); and using a PCA dimension reduction algorithm to perform data dimension reduction processing on the power-transformed matrix to obtain a data dimension reduction result for the audio data (104). The accuracy of PCA module data dimension reduction in the AVS encoder can be improved.

Description

Audio data processing method, device, chip and electronic device Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to an audio data processing method, device, chip, and electronic device. Background Art

[0002] With the increasing popularity of audio services, the next-generation Audio Video Coding Standard (AVS) encoder has been widely adopted. During the audio data encoding process, the AVS encoder employs a principal component analysis (PCA) data dimensionality reduction algorithm. This algorithm reduces the dimensionality of the Modulated Discrete Fourier Transform (MDFT) data matrix of the audio data, mapping it from a high-dimensional space to a low-dimensional space. This helps extract key information from the audio data and compresses the audio data size.

[0003] However, the PCA dimensionality reduction algorithm primarily considers the degree of variance explained when selecting principal components. In some cases, important principal component features with small variance may be ignored, resulting in low data dimensionality reduction accuracy in the PCA module of the AVS encoder.

[0004] Summary of the Invention

[0005] The present application provides an audio data processing method, device, chip and electronic device, which can use power transformation processing to enable the PCA module in the AVS encoder to select the principal component, so that the principal component matrix selected has a larger variance ratio when selecting the principal component, thereby solving the technical problem of low data dimensionality reduction accuracy of the PCA module in the AVS encoder.

[0006] A first embodiment of the present application discloses an audio data processing method, the method comprising:

[0007] Obtain a data matrix of audio data, where the data matrix is ​​an MDFT data matrix corresponding to the audio data in the AVS encoder;

[0008] Determining a target parameter value of a power operation parameter corresponding to each column vector in the data matrix;

[0009] Performing a power transformation on each column vector in the data matrix based on the target parameter value to obtain a power transformation matrix corresponding to the data matrix, wherein the power transformation is used to concentrate the variance of the data matrix in the direction of the principal component and enhance the feature proportion of the principal component in the data matrix;

[0010] The PCA dimensionality reduction algorithm is used to perform data dimensionality reduction processing on the power transformation matrix to obtain a data dimensionality reduction result of the audio data.

[0011] A second embodiment of the present application discloses an audio data processing device, the device comprising:

[0012] An acquisition module is used to acquire a data matrix of audio data, where the data matrix is ​​an MDFT data matrix corresponding to the audio data in the AVS encoder;

[0013] a determination module, configured to determine a target parameter value of a power operation parameter corresponding to each column vector in the data matrix;

[0014] A power transformation module is used to perform a power transformation on each column vector in the data matrix based on the target parameter value to obtain a power transformation matrix corresponding to the data matrix, wherein the power transformation is used to concentrate the variance of the data matrix in the direction of the principal component and enhance the feature proportion of the principal component in the data matrix;

[0015] The processing module is used to perform data dimensionality reduction processing on the power transformation matrix using the PCA dimensionality reduction algorithm to obtain a data dimensionality reduction result of the audio data.

[0016] The third aspect embodiment of the present application discloses an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method described in the first aspect embodiment of the present application.

[0017] The fourth aspect embodiment of the present application proposes a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the method described in the first aspect embodiment of the present application.

[0018] The fifth embodiment of the present application proposes a chip, which includes one or more interface circuits and one or more processors; the interface circuit is used to receive signals from the memory of the electronic device and send signals to the processor, the signals including computer instructions stored in the memory, and when the processor executes the computer instructions, the electronic device executes the method described in the embodiment of the first aspect of the present disclosure.

[0019] In summary, according to the audio data processing method, device, chip and electronic device proposed in the present disclosure, a data matrix of audio data can be obtained; then the target parameter value of the power operation parameter corresponding to each column vector in the data matrix is ​​determined; based on the target parameter value, each column vector in the data matrix is ​​subjected to power transformation to obtain a power transformation matrix corresponding to the data matrix, and the power transformation is used to concentrate the variance of the data matrix in the direction of the principal component and enhance the characteristic proportion of the principal component in the data matrix; finally, the power transformation matrix is ​​subjected to data dimensionality reduction processing using the PCA dimensionality reduction algorithm to obtain a data dimensionality reduction result of the audio data. The technical solution in the present disclosure can be subjected to power transformation processing to concentrate the variance of the data matrix corresponding to the audio data in the direction of the principal component, enhance the characteristic proportion of the principal component in the data matrix, and enable the PCA module in the AVS encoder to select the principal component, so that the principal component matrix selected has a larger variance proportion when selecting the principal component, thereby further improving the accuracy of the data dimensionality reduction of the PCA module.

[0020] Additional aspects and advantages of the present disclosure will be given in part in the description below and in part will be obvious from the description below, or will be learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute an improper limitation of the present disclosure.

[0022] FIG1 is a flow chart of an audio data processing method provided in an embodiment of the present application;

[0023] FIG2 is a flow chart of a method for calculating exponential operation parameters according to an embodiment of the present application;

[0024] FIG3 is a schematic diagram of an example of a data matrix provided in an embodiment of the present application;

[0025] FIG4 is a schematic diagram of an example of a restoration matrix corresponding to a PCA dimensionality reduction algorithm provided in an embodiment of the present application;

[0026] FIG5 is a schematic diagram of an example of a restoration matrix corresponding to a PCA dimensionality reduction algorithm after power transformation provided in an embodiment of the present application;

[0027] FIG6 is a schematic structural diagram of an audio data processing device provided in an embodiment of the present application;

[0028] FIG7 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] The following describes in detail embodiments of the present disclosure, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.

[0030] The embodiments of the present application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application. It should be noted that the embodiments of the present application and the features in the embodiments may be combined with each other unless there is a conflict.

[0031] The terms used in the embodiments of this application are for the purpose of describing specific embodiments only and are not intended to limit the embodiments of this application. The singular forms "a" and "the" used in the embodiments of this application and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0032] It should be understood that although the terms first, second, third, etc. may be used to describe various information in the embodiments of the present application, these information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of the embodiments of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0033] In order to solve the above technical problems, the present invention provides an audio data processing method, device, chip and electronic device, which can enable the PCA dimensionality reduction algorithm to accurately screen out the main components in the audio data, improve the accuracy of data dimensionality reduction in the AVS encoder, and enhance the effectiveness of the AVS encoder in encoding audio data.

[0034] The Principal Component Analysis (PCA) algorithm belongs to the field of multivariate statistical analysis. Multivariate statistical analysis is a statistical field that studies the relationships between multiple variables. The PCA algorithm, as a method within multivariate statistical analysis, is used to reduce data dimensionality and extract key features. Due to its advantages in feature extraction, dimensionality reduction, and data analysis, the PCA algorithm has applications in many different fields. For example, in data preprocessing, the PCA algorithm can be used to reduce data dimensionality and noise, improving the effectiveness of subsequent analysis. It can be used for tasks such as data cleaning, feature selection, and noise reduction. Image processing and computer vision: The PCA algorithm has a wide range of applications in these fields. It can be used for tasks such as image compression, feature extraction, face recognition, object detection, and image classification. Data visualization: The PCA algorithm can map high-dimensional data into a low-dimensional space, helping to visualize and understand the data's structure and relationships. It can be used for dataset visualization, cluster analysis, anomaly detection, and more. For example, the technical solution of the present invention can be applied to the compression processing of audio signals, mapping high-dimensional data into a low-dimensional space. This is applicable to a variety of business data types, without specific limitations here.

[0035] PCA is a commonly used data dimensionality reduction algorithm used to extract core features from a data matrix. It projects the original data onto a new feature space through a linear transformation, resulting in a principal component matrix. The original data information is concentrated in the principal components with the largest variance in the principal component matrix. Retaining the principal component vectors with the largest variance in the principal component matrix preserves most of the data information while reducing the data's dimensionality. However, PCA has limitations. When selecting principal components, PCA primarily considers the degree of variance explained. In some cases, important features with small variance may be overlooked, resulting in low data dimensionality reduction accuracy.

[0036] In view of this, as a possible implementation method, the power transformation model (Box-Cox) can be used to change the distribution of the original data to make it closer to a normal distribution. The calculation equation of the Box-Cox algorithm is shown as follows:

[0037] Where y is the column element in the column vector in the data matrix, and λ is the power operation parameter vector.

[0038] The Box-Cox algorithm aims to perform a Box-Cox power transformation on the elements in the column vectors of the input data matrix. The key feature of the Box-Cox power transformation is the introduction of a transformation parameter, λ, which is estimated from the data itself to determine the appropriate data transformation. Maximum likelihood estimation is used on the transformed data to calculate the value of the λ parameter that best approximates a normal distribution for the column vector. Once this value is calculated, the Box-Cox transformed output data is obtained. The Box-Cox algorithm improves the normality of vector data, making it more consistent with a normal distribution. Specifically, a Box-Cox transformation is performed on each column vector of the data matrix to obtain a data matrix with optimized normality. This data matrix is ​​then used as input to the PCA dimensionality reduction algorithm. The PCA dimensionality reduction algorithm is more effective when the data matrix conforms to a normal distribution.

[0039] However, while the Box-Cox algorithm can improve the normality of column vectors, it requires that all input data be positive. This is evident from the Box-Cox algorithm's equation, which significantly limits the data it can handle. Furthermore, the Box-Cox algorithm requires a large amount of data. If the number of elements in the column vector is insufficient, the algorithm may derive an incorrect transformation parameter λ, even worsening the normality of the column vector. For example, when the input data matrix is ​​a 4-dimensional square matrix, the column vector contains only 4 elements. The Box-Cox algorithm cannot effectively improve the normality of a vector with only 4 elements, and may even worsen the normality of the vector. This results in poor performance when processing data matrices with fewer column vectors. Although the Box-Cox algorithm encountered theoretical issues in experiments with 4-dimensional matrices (the Box-Cox algorithm cannot effectively improve the normality of vectors in 4-dimensional square matrices), experimental results indicate that applying exponential operations to the column vectors of a data matrix can affect the effectiveness of the PCA dimensionality reduction algorithm.

[0040] In view of this, as a possible implementation method, an improved PCA dimensionality reduction algorithm may be provided. FIG1 is a flowchart of an audio data processing method according to an exemplary embodiment. As shown in FIG1 , the method includes the following steps.

[0041] Step 101: Acquire a data matrix of audio data, where the data matrix is ​​an MDFT data matrix corresponding to the audio data in an AVS encoder.

[0042] Among them, the audio data can be the original audio data processed by a series of modules in the new generation audio and video coding standard (AVS) encoder, and then input into the discrete Fourier transform (MDFT) transformation module audio signal. The original audio data can be audio type data generated in any business scenario, and the business scenario for generating audio data is not specifically limited here. In the AVS encoder, the original audio data can be processed by a series of modules to obtain audio data, and then the audio data can be transformed by MDFT to obtain an MDFT data matrix. The MDFT data matrix can be used as the input of the subsequent PCA dimensionality reduction module in the AVS encoder for data dimensionality reduction processing. Data dimensionality reduction refers to reducing high-dimensional data to lower-dimensional data, which can remove redundant information in the data and retain the most important features, thereby reducing storage space and computing costs.

[0043] In the disclosed embodiments, the processing results of the original audio data by the corresponding preceding module of the PCA dimensionality reduction module in the AVS encoder are obtained to obtain a data matrix A corresponding to the audio data. In data matrix A, there are a total of m column vectors, and A = {a1, a2, a3…am} can be obtained. Each column in the data matrix is ​​treated as a set of column vectors and can be used as a feature to participate in the subsequent exponentiation process.

[0044] Step 102: Determine the target parameter value of the power operation parameter corresponding to each column vector in the data matrix.

[0045] In specific application scenarios, to improve the performance of the PCA dimensionality reduction algorithm, corresponding exponentiation parameters can be determined for different column vectors of the data matrix. Different column vectors correspond to different exponentiation parameters, and the elements in the same column vector are exponentiated using the same exponentiation parameter. In the disclosed embodiment, the target parameter value of the exponentiation parameter corresponding to each column vector is solved in order to find a set of exponentiation parameters corresponding to the data matrix. Under the power transformation of this set of exponentiation parameters, the variance of the data matrix can be most concentrated in the direction of the principal component, that is, the variance proportion of the principal component matrix selected in the PCA dimensionality reduction algorithm is increased.

[0046] In the embodiment of the present disclosure, when determining the target parameter value of the power operation parameter corresponding to each column vector in the data matrix, as shown in FIG2 , the embodiment steps may include:

[0047] Step 102-1: Determine a power operation parameter vector corresponding to the data matrix. The power operation parameter vector includes a power operation parameter for each column vector in the data matrix, where different column vectors correspond to different power operation parameters.

[0048] When the data matrix contains m column vectors, the power operation parameter vector corresponding to the determined data matrix may be λ={λ1,λ2,λ3...λm}.

[0049] Step 102-2: Determine an initial power function matrix corresponding to the data matrix based on the initial power function equation and the power operation parameter vector.

[0050] Among them, the formula characteristics of the initial power function equation are described as:

[0051] Where y is the column element in the column vector, |y (λ) | is the absolute value of the power operation result corresponding to the column element, y λ The positive and negative signs of are the same as those of y, λ is the power operation parameter vector, λ={λ1,λ2,λ3...λm}, and the power operation parameter vector may include the power operation parameters λ1,λ2,λ3...,λm corresponding to each column vector in the data matrix, where m is the number of column vectors.

[0052] For the embodiments of the present disclosure, compared with the limitation of the Box-Cox algorithm that the calculation equation requires that the input data must all be positive numbers, the power operation equation is optimized and improved in this scheme. By solving the absolute value of the power operation result corresponding to the column elements, it is able to process non-positive elements.

[0053] Accordingly, when determining the initial power function matrix corresponding to the data matrix based on the initial power function equation and the power operation parameter vector, the power operation parameter corresponding to each column vector can be substituted into the initial power function equation to obtain the power function equation corresponding to the column vector containing the unknown number λ, and then the power operation result of each column element under the column vector is calculated based on the power function equation. Exemplarily, for the first column vector in the data matrix, after substituting the corresponding power operation parameter λ1 into the initial power function equation, the power function equation corresponding to the first column vector containing the unknown number λ1 can be obtained: Based on this power function equation, we can calculate the power operation result of each column element under the first column vector containing the unknown number λ1; for the second column vector in the data matrix, after substituting the corresponding power operation parameter λ2 into the initial power function equation, we can obtain the power function equation of the second column vector containing the unknown number λ2: Then, based on this power function equation, we calculate the power operation result of each column element under the second column vector containing the unknown number λ2; ... For the mth column vector in the data matrix, after substituting the corresponding power operation parameter λm into the initial power function equation, we can obtain the power function equation of the first column vector containing the unknown number λm: Then, based on the power function equation, the power operation result of each column element under the mth column vector containing the unknown number λm is calculated. After the m column elements of the data matrix are all calculated by the above power operation results, the initial power function matrix B corresponding to the data matrix can be obtained. λ1 ,a2 λ2 ,a3 λ3 ...am λm}.

[0054] Step 102-3: Calculate the first covariance matrix of the initial power function matrix, perform eigenvalue decomposition on the first covariance matrix, and obtain multiple column vectors corresponding to multiple first eigenvalues ​​containing the power operation parameter vector. The first eigenvalues ​​are used to characterize the relationship between the variance distribution of the eigenvectors of the data matrix and the change of the power operation parameter vector in the power transformation.

[0055] In specific application scenarios, the variance distribution of the data matrix is ​​reflected in the eigenvalue of its covariance matrix. The larger the eigenvalue, the greater the variance on its eigenvector. At this time, the variance distribution changes with the power operation parameter vector λ, so the first covariance matrix B of the initial power function matrix B can be solved T B, by the first covariance matrix B T B performs eigenvalue decomposition to obtain multiple column vectors corresponding to multiple first eigenvalues ​​of the power operation parameter vector {eval1 λ , eval2 λ , eval3 λ ...evalm λ Based on multiple first eigenvalues, the optimal exponentiation parameters corresponding to each column vector are found. This increases the variance in the direction of the principal component, concentrating the variance of the data matrix in the direction of the principal component and enhancing the characteristic proportion of the principal component in the data matrix. Furthermore, the eigenvectors corresponding to the k largest eigenvalues ​​can be accurately selected from multiple eigenvectors as principal components. The original data is projected onto the selected principal components to obtain the reduced-dimensional data. This allows the reduced-dimensional data to reflect the principal component characteristics of the original data, thereby increasing the variance proportion of the principal component matrix selected in the PCA dimensionality reduction algorithm, thereby improving the effectiveness of data dimensionality reduction.

[0056] Step 102 - 4 : Calculate a PCA performance index based on the multiple first eigenvalues, and determine a target parameter value of a power operation parameter corresponding to each column vector according to the PCA performance index.

[0057] In a specific application scenario, when determining the optimal power operation parameters corresponding to each column vector, that is, determining the target parameter values of the power operation parameters corresponding to each column vector, a PCA performance metric L can be defined. By calculating the PCA performance metric L, the quality of the results of the PCA dimensionality reduction algorithm for the power transformation matrix under different power operation parameter vectors λ can be compared. The larger the value of the PCA performance metric L, the better the results of the PCA dimensionality reduction algorithm for the data matrix obtained by the power operation. Specifically, a PCA performance metric estimation equation for evaluating the PCA performance metric L based on the power operation parameter vector λ can be provided. Then, based on the PCA performance metric estimation equation and multiple first eigenvalues, the following calculation process can be repeatedly executed until it is determined that the PCA performance metric reaches the maximum value. When the PCA performance metric reaches the maximum value, extract the power operation parameters corresponding to each column vector: input the power operation parameter vector into the PCA performance metric estimation equation, and select a preset number of first eigenvalues from the multiple first eigenvalues and input them into the PCA performance metric estimation equation to calculate the PCA performance metric.

[0058] Among them, the formula characteristics of the PCA performance metric estimation equation are described as follows:

[0059] In the formula, L λ is the PCA performance metric, eval1 λ , eval1 λ ,..., evalk λ are k preset first eigenvalues among the multiple first eigenvalues, where k is the target dimension of data dimensionality reduction, k < m, evalsum is the sum of the absolute values of the multiple first eigenvalues, λ = {λ1, λ2, λ3... λm} is the power operation parameter vector, λ1, λ2, λ3..., λm are the power operation parameters corresponding to each column vector in the power operation parameter vector, and m is the number of column vectors.

[0060] Correspondingly, when inputting the power operation parameter vector into the PCA performance metric estimation equation and selecting a preset number of first eigenvalues from the multiple first eigenvalues and inputting them into the PCA performance metric estimation equation to calculate the PCA performance metric, the PCA performance metric L can be regarded as an m-dimensional equation about m power operation parameters. The larger the value of the PCA performance metric L, the better the results of the PCA dimensionality reduction algorithm for the data matrix obtained by the power operation. Therefore, the λ vector corresponding to the maximum value of the above PCA performance metric estimation equation is the optimal power operation parameter vector. The optimal power operation parameter vector may include the specific parameter values, that is, the target parameter values, of the power operation parameters corresponding to each column vector.

[0061] In the embodiment of the present disclosure, the PCA performance index L is used to evaluate the quality of the data matrix power operation result. The calculation equation of the PCA performance index L does not rely on the normal performance generation after the power operation of a single column vector, but relies on whether the entire data matrix is ​​more suitable for processing by the PCA dimensionality reduction algorithm after the power operation. Determining the power transformation matrix through the PCA performance index L can not only overcome the problem of insufficient number of elements, but also make the transformed matrix more suitable for PCA dimensionality reduction.

[0062] Step 103: Perform a power transformation on each column vector in the data matrix based on the target parameter value to obtain a power transformation matrix corresponding to the data matrix. The power transformation is used to concentrate the variance of the data matrix in the direction of the principal component and enhance the feature proportion of the principal component in the data matrix.

[0063] For the embodiment of the present disclosure, when performing a power transformation on each column vector in the data matrix and determining the power transformation matrix corresponding to the data matrix, the target parameter value of the power operation parameter corresponding to each column vector can be substituted into the initial power function equation to calculate the power operation result corresponding to each column element; after the power operation result is calculated for each column element, the power transformation matrix corresponding to the data matrix is ​​obtained. Among them, the value range of the target parameter value can be (-2, 2). For example, if the target parameter value of the power operation parameter λ1 corresponding to the first column vector in the data matrix is ​​calculated to be 0, then the target parameter value can be substituted into the power function equation of the unknown number λ1: Based on the power function equation, we can calculate the power operation result corresponding to each column element under the first column vector as |y (λ1) |=ln(|y|+1); if the target parameter value corresponding to the power operation parameter data λ2 of the first column vector in the data matrix is ​​calculated to be 1, then this target parameter value can be substituted into the power function equation containing the unknown number λ2: Then, based on the power function equation, the power operation result corresponding to each column element under the second column vector is calculated as |y λ2 |=|y| 1 ; ..., after all the m column elements of the data matrix are calculated by the above power operation results, the power transformation matrix corresponding to the data matrix can be obtained.

[0064] Step 104: Use the PCA dimensionality reduction algorithm to perform data dimensionality reduction processing on the power transformation matrix to obtain a data dimensionality reduction result of the audio data.

[0065] In the disclosed embodiment, the power transformation matrix obtained in step 103 of the embodiment can be used as input to the PCA dimensionality reduction algorithm for PCA dimensionality reduction calculations. In the PCA dimensionality reduction algorithm, when restoring the original matrix, a corresponding inverse transformation can be performed based on the power operation parameter vector λ and the power function operation equation to restore the original audio data matrix. Given that the variance in the power transformation matrix is ​​concentrated in the direction of the principal component and the characteristics of the principal component account for a large proportion, the data dimensionality reduction effect of the audio data can be guaranteed, while also ensuring the recovery effect when restoring the audio data matrix.

[0066] In the PCA dimensionality reduction algorithm, the principal component matrix is ​​composed of the eigenvectors of the covariance matrix. It is an orthogonal matrix in which each column corresponds to a principal component. The column vectors of the principal component matrix are the eigenvectors of the covariance matrix, which are sorted according to the size of the corresponding eigenvalues. The variance distribution is reflected in the eigenvalues ​​of its covariance matrix. The variance calculation is the sum of the selected principal component eigenvalues ​​divided by the total sum of the eigenvalues. Therefore, the larger the eigenvalue, the greater the variance on its eigenvector. Assume that the power transformation matrix X is obtained through the steps of the above embodiment, wherein each column vector in the power transformation matrix X represents a feature and each row represents a sample. The principal component has a larger variance share in the power transformation matrix X. For the embodiment of the present disclosure, when the original data is subjected to data dimensionality reduction based on the power transformation matrix X: first, the covariance matrix C of the power transformation matrix X can be calculated. Then, C is subjected to eigenvalue decomposition to obtain the eigenvalues ​​and corresponding eigenvectors. The eigenvectors are arranged in descending order of eigenvalues, and the first k eigenvectors (k is the number of principal components selected, i.e., the target dimension of data dimensionality reduction) are selected to form the principal component matrix P. The dimensions of P are n × k, where n is the number of features in the original data. The column vectors of the principal component matrix P are the principal components of the data, which are linear combinations of the original data. Multiplying the power transformation matrix X corresponding to the original data with the principal component matrix P yields a projection into the principal component space, mapping the original data onto the principal components. This process can be expressed as Y = XP, where Y is the mapping matrix.

[0067] Correspondingly, for the embodiments of the present disclosure, when using the PCA dimensionality reduction algorithm to perform dimensionality reduction processing on the power transformation matrix to obtain the dimensionality reduction result of the audio data, the steps of the embodiments may include: performing normalization processing on the power transformation matrix so that the mean of each column vector in the power transformation matrix is 0 and the variance is 1. This step is to ensure that each feature has the same importance to the result of PCA; based on the normalized power transformation matrix, calculating the second covariance matrix between features, where the second covariance matrix describes the linear relationship between features; performing eigenvalue decomposition on the second covariance matrix to obtain multiple second eigenvalues corresponding to multiple column vectors and multiple second eigenvectors corresponding to the multiple second eigenvalues. The second eigenvalue is used to characterize the variance size of the second eigenvector after the data matrix undergoes power transformation, and the second eigenvector is used to represent the direction of the audio data in the new feature space; selecting a preset number k of second eigenvectors from the multiple second eigenvectors to form a principal component matrix in the order of the second eigenvalues from large to small, where k is the target dimension to which it is desired to reduce, and k < m; projecting the audio data onto the principal component matrix to obtain the dimensionality reduction result of the audio data.

[0068] Among them, in the PCA dimensionality reduction algorithm, the calculation of the second covariance matrix and the solution of the second eigenvector are because the core of the PCA dimensionality reduction algorithm is to find several directions with the most concentrated variance in the data matrix as the principal component matrix. These directions are reflected in the eigenvectors of the covariance matrix. The larger the eigenvalue of the eigenvector, the larger the variance in that direction. This step is to obtain k eigenvectors with the largest eigenvalues.

[0069] In a specific application scenario, through testing, the result of restoring the data matrix after power transformation processing is better when the PCA dimensionality reduction algorithm retains the same number of principal component components.

[0070] Taking a four-dimensional square matrix as the input matrix of the PCA dimensionality reduction algorithm (i.e., the MDFT data matrix corresponding to the audio data), and it is required to retain two principal component components as a test case. The MDFT data matrix can be seen in Figure 3, the restoration matrix corresponding to the PCA dimensionality reduction algorithm can be seen in Figure 4, and the restoration matrix corresponding to the PCA dimensionality reduction algorithm after power transformation can be seen in Figure 5. Through comparison, it can be obtained that the restoration matrix of the PCA dimensionality reduction algorithm improved by the power operation is closer to the original matrix. The sum of the relative errors is 0.0164972 (the sum of the absolute values of the differences between the corresponding elements divided by the original elements), and the sum of the relative errors of the unimproved PCA dimensionality reduction algorithm is 11.8591. It can be seen that the technical solution in the present disclosure can effectively perform dimensionality reduction and improve the accuracy of dimensionality reduction.

[0071] In summary, according to the audio data processing method proposed in the present disclosure, a data matrix of audio data can be obtained; then, the power operation parameters corresponding to each column vector in the data matrix are determined; each column vector in the data matrix is ​​subjected to power transformation based on the target parameter value to obtain a power transformation matrix corresponding to the data matrix, and the power transformation is used to concentrate the variance of the data matrix in the direction of the principal component and enhance the characteristic proportion of the principal component in the data matrix; finally, the power transformation matrix is ​​subjected to data dimensionality reduction processing using the PCA dimensionality reduction algorithm to obtain a data dimensionality reduction result of the audio data. The technical solution in the present disclosure can be subjected to power transformation processing to concentrate the variance of the data matrix corresponding to the audio data in the direction of the principal component, enhance the characteristic proportion of the principal component in the data matrix, and enable the PCA module in the AVS encoder to select the principal component, so that the principal component matrix selected has a larger variance proportion when selecting the principal component, thereby further improving the accuracy of the data dimensionality reduction of the PCA module.

[0072] Based on the implementation of the method shown in FIG1 , this embodiment provides an audio data processing device 600 . As shown in FIG6 , the device 600 includes: a generation module 601 , a determination module 602 , a power conversion module 603 , and a processing module 604 .

[0073] An acquisition module 601 is configured to acquire a data matrix of audio data, where the data matrix is ​​an MDFT data matrix corresponding to the audio data in an AVS encoder;

[0074] A determination module 602 may be configured to determine a target parameter value of a power operation parameter corresponding to each column vector in the data matrix;

[0075] A power transformation module 603 may be used to perform a power transformation on each column vector in the data matrix based on the target parameter value to obtain a power transformation matrix corresponding to the data matrix. The power transformation is used to concentrate the variance of the data matrix in the direction of the principal component and enhance the feature proportion of the principal component in the data matrix.

[0076] The processing module 604 may be configured to perform data dimensionality reduction processing on the power transformation matrix using a PCA dimensionality reduction algorithm to obtain a data dimensionality reduction result of the audio data.

[0077] In one embodiment of the present application, the determination module 602 can be specifically used to determine the power operation parameter vector λ corresponding to the data matrix, where the power operation parameter vector λ includes the power operation parameters of each column vector in the data matrix, wherein different column vectors correspond to different power operation parameters; based on the initial power function equation and the power operation parameter vector λ, the initial power function matrix corresponding to the data matrix is ​​determined; the first covariance matrix of the initial power function matrix is ​​calculated; the first covariance matrix is ​​subjected to eigenvalue decomposition to obtain multiple first eigenvalues ​​corresponding to multiple column vectors including the power operation parameter vector λ, the first eigenvalues ​​being used to characterize the relationship between the variance distribution of the eigenvectors of the data matrix and the change in the power operation parameter vector λ during the power transformation; the PCA performance index is calculated based on the multiple first eigenvalues, and the target parameter value of the power operation parameter corresponding to each column vector is determined according to the PCA performance index.

[0078] In one embodiment of the present application, the formula characteristics of the initial power function equation are described as follows:

[0079] Where y is the column element in the column vector, |y (λ) | is the absolute value of the power operation result corresponding to the column element, y λ The sign of is the same as that of y, and λ is the power operation parameter vector.

[0080] In one embodiment of the present application, the determination module 602 can be specifically used to repeatedly execute the following calculation process until it is determined that the PCA performance index reaches a maximum value, and extract the target parameter value of the power operation parameter corresponding to each column vector when the PCA performance index reaches the maximum value: input the power operation parameter vector into the PCA performance index estimation equation, and select a preset number of first eigenvalues ​​from multiple first eigenvalues ​​and input them into the PCA performance index estimation equation to calculate the PCA performance index.

[0081] In one embodiment of the present application, the PCA performance index estimation equation is characterized by the formula described as:

[0082] Where, L λ is the PCA performance index, eval1 λ ,eval1 λ ,...,evalk λ is a preset number of first eigenvalues ​​among multiple first eigenvalues, the preset number is the target dimension of data dimensionality reduction, evalsum is the sum of the absolute values ​​of the multiple first eigenvalues, λ = {λ1,λ2,λ3...λm} is the power operation parameter vector, λ1,λ2,λ3...,λm are the power operation parameters corresponding to each column vector in the power operation parameter vector, and m is the number of column vectors.

[0083] In one embodiment of the present application, the power transformation module 603 can be specifically used to substitute the target parameter value of the power operation parameter corresponding to each column vector into the initial power function equation, and calculate the power operation result corresponding to each column element; after the power operation result is calculated for each column element, the power transformation matrix corresponding to the data matrix is ​​obtained.

[0084] In one embodiment of the present application, the processing module 604 can be used to standardize the power transformation matrix so that the mean of each column vector in the power transformation matrix is ​​0 and the variance is 1; based on the standardized power transformation matrix, the second covariance matrix between the column vectors is calculated; the second covariance matrix is ​​decomposed into eigenvalues ​​to obtain multiple second eigenvalues ​​corresponding to the multiple column vectors, and multiple second eigenvectors corresponding to the multiple second eigenvalues, and the second eigenvalues ​​are used to characterize the variance of the second eigenvectors after the data matrix is ​​power transformed; in descending order of the second eigenvalues, a preset number of second eigenvectors are selected from the multiple second eigenvectors to form a principal component matrix; the audio data is projected onto the principal component matrix to obtain the data dimensionality reduction result of the audio data.

[0085] FIG7 is a block diagram of an electronic device 800 for implementing the above audio data processing method according to an exemplary embodiment.

[0086] For example, the electronic device 800 may be a mobile phone, a computer, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0087] 7 , the electronic device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0088] The processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to perform all or part of the steps of the above-described method. In addition, the processing component 802 may include one or more modules to facilitate interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.

[0089] The memory 804 is configured to store various types of data to support operations on the electronic device 800. Examples of such data include instructions for any application or method operating on the electronic device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0090] The power supply component 806 provides power to the various components of the electronic device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 800.

[0091] The multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor can not only sense the boundaries of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the electronic device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have a focal length and optical zoom capability.

[0092] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), which is configured to receive external audio signals when the electronic device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in the memory 804 or transmitted via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0093] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as a keyboard, click wheel, buttons, etc. These buttons may include but are not limited to: a home button, volume buttons, a start button, and a lock button.

[0094] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the electronic device 800. For example, the sensor assembly 814 can detect the open / closed state of the electronic device 800, the relative positioning of components, such as the display and keypad of the electronic device 800. The sensor assembly 814 can also detect changes in the position of the electronic device 800 or a component of the electronic device 800, the presence or absence of user contact with the electronic device 800, the orientation or acceleration / deceleration of the electronic device 800, and temperature changes of the electronic device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0095] The communication component 816 is configured to facilitate wired or wireless communication between the electronic device 800 and other devices. The electronic device 800 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, 4G LTE, 5G NR (New Radio) or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0096] In an exemplary embodiment, the electronic device 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above methods.

[0097] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, and the instructions can be executed by the processor 820 of the electronic device 800 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0098] The embodiments of the present disclosure further provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute the audio data processing method described in the above embodiments of the present disclosure.

[0099] An embodiment of the present disclosure also proposes a chip, which includes one or more interface circuits and one or more processors; the interface circuit is used to receive signals from the memory of the electronic device and send signals to the processor, the signals including computer instructions stored in the memory, and when the processor executes the computer instructions, the electronic device executes the audio data processing method described in the above embodiments of the present disclosure.

[0100] It should be noted that the terms "first," "second," and the like in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the numbers used in this manner are interchangeable where appropriate so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure as detailed in the appended claims.

[0101] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with an embodiment or example is included in at least one embodiment or example of the present disclosure. In this specification, the illustrative use of the above terms does not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0102] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present disclosure includes additional implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present disclosure belong.

[0103] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processing module, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection having one or more wires (control method), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or otherwise processing it in a suitable manner if necessary, and then storing it in a computer memory.

[0104] It should be understood that the various parts of the embodiments of the present disclosure can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0105] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0106] In addition, the functional units in the various embodiments of the present disclosure may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into a module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The above-mentioned storage medium may be a read-only memory, a magnetic disk, an optical disk, etc.

[0107] Although the embodiments of the present disclosure have been shown and described above, it is understood that the above embodiments are exemplary and are not to be construed as limitations on the present disclosure. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present disclosure.

Claims

1. An audio data processing method, characterized in that: include: Acquire a data matrix of audio data, wherein the data matrix is ​​an MDFT data matrix corresponding to the audio data in the AVS encoder; Determine a target parameter value of a power operation parameter corresponding to each column vector in the data matrix; Performing a power transformation on each column vector in the data matrix based on the target parameter value to obtain a power transformation matrix corresponding to the data matrix, wherein the power transformation is used to concentrate the variance of the data matrix in the direction of the principal component and enhance the characteristic proportion of the principal component in the data matrix; The PCA dimensionality reduction algorithm is used to perform data dimensionality reduction processing on the power transformation matrix to obtain a data dimensionality reduction result of the audio data.

2. The method according to claim 1, characterized in that The step of determining a target parameter value of a power operation parameter corresponding to each column vector in the data matrix comprises: Determine a power operation parameter vector λ corresponding to the data matrix, wherein the power operation parameter vector λ includes a power operation parameter of each column vector in the data matrix, wherein different column vectors correspond to different power operation parameters; Determine an initial power function matrix corresponding to the data matrix based on an initial power function equation and the power operation parameter vector λ; Calculating a first covariance matrix of the initial power function matrix; Performing eigenvalue decomposition on the first covariance matrix to obtain a plurality of first eigenvalues ​​corresponding to a plurality of column vectors containing the power operation parameter vector λ, wherein the first eigenvalues ​​are used to characterize a relationship between a variance distribution of an eigenvector of the data matrix and a change in the power operation parameter vector λ during a power transformation; A PCA performance index is calculated based on the multiple first eigenvalues, and a target parameter value of the power operation parameter corresponding to each column vector is determined according to the PCA performance index.

3. The method according to claim 2, characterized in that The formula characteristic description of the initial power function equation is: In the formula, y is the column element in the column vector, |y (λ) | is the absolute value of the power operation result corresponding to the column element, y λ The positive and negative signs of are the same as those of y, and λ is the power operation parameter vector.

4. The method according to claim 2, characterized in that: The calculating of the PCA performance index based on the multiple first eigenvalues, and determining the target parameter value of the power operation parameter corresponding to each column vector according to the PCA performance index, comprises: The following calculation process is repeatedly performed until it is determined that the PCA performance index reaches a maximum value, and the target parameter value of each column vector corresponding to the power operation parameter when the PCA performance index reaches the maximum value is extracted: The power operation parameter vector is input into the PCA performance index estimation equation, and a preset number of first eigenvalues ​​are selected from the multiple first eigenvalues ​​and input into the PCA performance index estimation equation to calculate the PCA performance index.

5. The method according to claim 4, characterized in that The formula characteristic description of the PCA performance index estimation equation is: Where, L λ is the PCA performance index, eval1 λ ,eval1 λ ,…,evalk λ is a preset number of first eigenvalues ​​among the multiple first eigenvalues, the preset number is the target dimension of data dimensionality reduction, evalsum is the sum of the absolute values ​​of the multiple first eigenvalues, λ={λ1,λ2,λ3…λm} is the power operation parameter vector, λ1, λ2,λ3…,λm are the power operation parameters corresponding to each column vector in the power operation parameter vector, and m is the number of column vectors.

6. The method according to claim 2, characterized in that The performing a power transformation on each column vector in the data matrix based on the target parameter value to obtain a power transformation matrix corresponding to the data matrix includes: Substitute the target parameter value of the power operation parameter corresponding to each column vector into the initial power function equation, and calculate the power operation result corresponding to each column element; After the power operation result is calculated for each column element, a power transformation matrix corresponding to the data matrix is ​​obtained.

7. The method according to claim 1, characterized in that The step of performing data dimensionality reduction processing on the power transformation matrix using the PCA dimensionality reduction algorithm to obtain a data dimensionality reduction result of the audio data includes: Standardizing the power transformation matrix so that each column vector in the power transformation matrix has a mean of 0 and a variance of 1; Based on the normalized power transformation matrix, calculate the second covariance matrix between the column vectors; Performing eigenvalue decomposition on the second covariance matrix to obtain a plurality of second eigenvalues ​​corresponding to a plurality of column vectors and a plurality of second eigenvectors corresponding to the plurality of second eigenvalues, wherein the second eigenvalues ​​are used to characterize the variance of the second eigenvectors of the data matrix after power transformation; In the descending order of the second eigenvalues, a preset number of second eigenvectors are selected from the plurality of second eigenvectors to form a principal component matrix; The audio data is projected onto the principal component matrix to obtain a data dimensionality reduction result of the audio data.

8. An audio data processing device, characterized in that: include: An acquisition module, used to acquire a data matrix of audio data, wherein the data matrix is ​​an MDFT data matrix corresponding to the audio data in the AVS encoder; A determination module, used to determine a target parameter value of a power operation parameter corresponding to each column vector in the data matrix; A power transformation module, used for performing a power transformation on each column vector in the data matrix based on the target parameter value to obtain a power transformation matrix corresponding to the data matrix, wherein the power transformation is used to concentrate the variance of the data matrix in the direction of the principal component and enhance the characteristic proportion of the principal component in the data matrix; The processing module is used to perform data dimensionality reduction processing on the power transformation matrix by using the PCA dimensionality reduction algorithm to obtain the data dimensionality reduction result of the audio data.

9. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-7.

11. A chip, characterized in that: It comprises one or more interface circuits and one or more processors; the interface circuit is used to receive a signal from a memory of an electronic device and send the signal to the processor, the signal includes a computer instruction stored in the memory, and when the processor executes the computer instruction, the electronic device executes the method described in any one of claims 1 to 7.

Images