A method, device and medium for active noise reduction of multiple sources under complex working conditions

Abstract

The present invention discloses a method, device and medium for active noise reduction of multiple sources under complex working conditions, the method comprising: collecting the in-vehicle noise signal of the target vehicle; frequency locking the main component characteristic signal with the largest sparsity in the in-vehicle noise signal to obtain a digital signal; generating an active noise reduction signal with the same signal peak value and opposite phase as the digital signal, and outputting the active noise reduction signal to the target vehicle. The present invention proposes a method, device and medium for active noise reduction of multiple sources under complex working conditions, by frequency locking the main component characteristic signal with the largest sparsity and generating an active noise reduction signal with the opposite phase, the most important noise component can be targeted for noise reduction processing, which can solve the problem of poor adaptability of the active noise reduction method in the vehicle.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle noise reduction, and in particular to a method, device and medium for active noise reduction of multiple sources under complex working conditions. Background Art

[0002] According to the source, the noise signals inside the car are mainly divided into three categories: powertrain noise, road noise and wind noise. Excessive noise signals inside the car will distract the driver's attention and reduce driving comfort; and long-term exposure to high-decibel noise environment may cause hearing damage to passengers and drivers; through effective noise reduction measures, a quieter driving environment can be created, allowing drivers to focus more on road conditions and driving operations, thereby improving driving safety and stability. Good noise reduction effects can also significantly improve customer satisfaction. Conventional active noise reduction methods mainly rely on the minimum error value method (LMS) of the input and output channels for adjustment. The LMS algorithm is an adaptive filter algorithm based on the gradient descent method. Its core idea is to update the filter weights by minimizing the mean square error of the error signal, thereby gradually approaching the optimal filter.

[0003] However, in practical applications, vehicle operating conditions are complex and the excitation source input is constantly changing, resulting in the inability of existing active noise reduction methods to match the driving scenes inside and outside the vehicle, and there are problems such as signal lag and poor adaptability. Summary of the invention

[0004] The present invention provides a multi-source noise active noise reduction method, device and medium under complex working conditions, so as to solve the problem of poor adaptability of the active noise reduction method in a vehicle.

[0005] Collect the interior noise signal of the target vehicle;

[0006] Frequency locking is performed on the main component characteristic signal with the largest sparsity in the in-vehicle noise signal to obtain a digital signal;

[0007] An active noise reduction signal having a signal peak value equal to that of the digital signal and an opposite phase is generated, and the active noise reduction signal is output to the target vehicle.

[0008] The present invention frequency locks the main component characteristic signal with the largest sparsity in the in-vehicle noise signal, which means that the most important and representative component in the in-vehicle noise signal can be accurately identified. The digital signal obtained by frequency locking is an important basis for reflecting the characteristics of the in-vehicle noise signal in real time; therefore, based on this digital signal, a corresponding active noise reduction signal can be quickly generated. The signal peak of this noise reduction signal is equal to the signal peak of the digital signal, but the phase is opposite, thereby ensuring the maximization of the noise reduction effect. In addition, the frequency locking technology can track the changes of the in-vehicle noise signal in real time. When the noise characteristics change, the frequency locking parameters can be quickly adjusted to ensure that the noise reduction signal always maintains the best phase and amplitude match with the noise signal. This real-time adjustment capability significantly enhances the adaptability of the present solution, enabling it to maintain a stable noise reduction effect in different noise environments.

[0009] Compared with the prior art, the present invention can perform noise reduction processing on the most important noise component in a targeted manner by frequency locking the main component characteristic signal with the greatest sparsity and generating an active noise reduction signal with the opposite phase thereto, thereby solving the problem of poor adaptability of the active noise reduction method in the car.

[0010] As a preferred solution, the frequency of the main component characteristic signal with the largest sparsity in the in-vehicle noise signal is locked to obtain a digital signal, which is specifically:

[0011] Adaptively filtering the in-vehicle noise signal to obtain a first output signal;

[0012] Performing sparse processing on the first output signal to obtain the principal component feature signal with the maximum sparsity;

[0013] If the error value between the principal component characteristic signal and the first output signal does not meet the preset error requirement, the parameters of the sparse processing are readjusted, and the principal component characteristic signal is iteratively processed until the iterative processing result meets the error requirement, thereby obtaining a second output signal;

[0014] Perform analog-to-digital conversion on the first n principal component feature signals with the largest sparsity in the second output signal to obtain the digital signal.

[0015] In this preferred scheme, if the error value between the main component characteristic signal and the first output signal does not meet the preset error requirement, the parameters of the sparse processing are readjusted and iterative processing is performed, so that the optimal solution can be gradually approached, the iteration efficiency can be improved, and the effectiveness of the output signal can be ensured.

[0016] As a preferred solution, the first output signal is subjected to sparse processing to obtain the principal component characteristic signal with the largest sparsity, specifically:

[0017] Calculating a covariance matrix between features in the first output signal, and performing eigenvalue set decomposition on the covariance matrix to obtain an eigenvalue set and a corresponding eigenvector set;

[0018] Establishing a coordinate system according to the first k largest eigenvalues ​​in the eigenvalue set and their corresponding eigenvectors in the eigenvector set to obtain a first coordinate system;

[0019] The first output signal is projected into the first coordinate system to obtain the principal component characteristic signal.

[0020] This preferred solution can capture the correlation between features by calculating the covariance matrix between the features in the output signal. Since the eigenvalues ​​and eigenvectors are calculated based on the covariance matrix, and the covariance matrix is ​​a stable statistic, the constructed coordinate system also has good stability, which is helpful for subsequent signal processing and data analysis. By projecting the output signal into the first coordinate system, the principal component characteristic signal can be obtained. This method realizes the dimensionality reduction of the signal, converting high-dimensional data into a low-dimensional representation, while retaining the most important information.

[0021] As a preferred solution, analog-to-digital conversion is performed on the first n principal component feature signals with the largest sparsity in the second output signal to obtain the digital signal, specifically:

[0022] Extracting the first n principal component characteristic signals with the largest sparsity in the second output signal by frequency locking, and treating the extracted signals as the first n-order noise frequencies;

[0023] Perform analog-to-digital conversion on the first n-order noise frequencies to obtain the digital signal.

[0024] This preferred solution uses frequency locking technology to accurately lock and extract the main component characteristic signal in the second output signal. This method can accurately identify and separate the key components in the signal. The extracted main component characteristic signal is regarded as the first n-order noise frequency, which helps to identify and distinguish the noise component and useful information in the signal, thereby ensuring that the subsequent converted signal is clear and stable.

[0025] As a preferred solution, an active noise reduction signal having a signal peak value equal to that of the digital signal and an opposite phase is generated, specifically:

[0026] Generating a sound wave signal having a signal peak value equal to that of the digital signal and an opposite phase, and performing digital-to-analog conversion on the sound wave signal to obtain an analog noise reduction signal;

[0027] The active noise reduction signal is obtained by reducing the noise reduction error of the analog noise reduction signal according to the weight memory feedback and adaptive active noise reduction system; wherein the weight of the weight memory feedback and adaptive active noise reduction system is updated according to the analog noise reduction signal and the vehicle body vibration signal, and the noise reduction error is calculated according to the in-vehicle noise signal and the analog noise reduction signal.

[0028] This preferred solution can accurately produce a noise reduction signal that counteracts noise by generating a sound wave signal that is equal to the signal peak value and opposite in phase to the digital signal. This method is based on the physical properties of sound waves and ensures that the noise reduction signal and the noise are completely opposite in phase, thereby achieving effective noise cancellation.

[0029] As a preferred solution, the weights of the weight memory feedback and adaptive active noise reduction system are updated according to the simulated noise reduction signal and the vehicle body vibration signal, specifically:

[0030] If the noise reduction error of the analog noise reduction signal is greater than a first preset value, a first weight is calculated according to the first formula, and the weight memory feedback and the weight in the adaptive active noise reduction system are updated to the first weight;

[0031] Among them, the first formula is:

[0032] F=sigmoid{(w(n)*x(n)+b)*tanh(w(n)*x(n)+b)}

[0033] Among them, w(n) is the input signal weight at the current moment, x(n) is the vehicle body vibration signal, b is a constant, and sigmoid and tanh are activation functions.

[0034] In practical applications, the noise environment of this preferred solution is often complex and changeable. By dynamically adjusting the weights according to the noise reduction error, the current noise characteristics can be optimized, thereby improving the accuracy and effect of noise reduction.

[0035] The present application also provides a multi-source noise active noise reduction device under complex working conditions, including a collection module, a locking module and a noise reduction module;

[0036] Wherein, the acquisition module is used to collect the interior noise signal of the target vehicle;

[0037] The locking module is used to perform frequency locking on the main component characteristic signal with the largest sparsity in the in-vehicle noise signal to obtain a digital signal;

[0038] The noise reduction module is used to generate an active noise reduction signal having a signal peak value equal to that of the digital signal and an opposite phase, and output the active noise reduction signal to the target vehicle.

[0039] As a preferred solution, the locking module includes a filtering unit, a sparse unit, an error unit and a conversion unit;

[0040] The filtering unit is used to adaptively filter the in-vehicle noise signal to obtain a first output signal;

[0041] The sparseness unit is used to perform sparseness processing on the first output signal to obtain the principal component feature signal with the maximum sparsity;

[0042] The error unit is used to readjust the parameters of the sparse processing if the error value between the principal component feature signal and the first output signal does not meet the preset error requirement, and iteratively process the principal component feature signal until the iterative processing result meets the error requirement to obtain the second output signal;

[0043] The conversion unit is used to perform analog-to-digital conversion on the first n principal component feature signals with the largest sparsity in the second output signal to obtain the digital signal.

[0044] As a preferred solution, the sparseness unit includes a decomposition subunit, a coordinate subunit and a projection subunit;

[0045] The decomposition subunit is used to calculate the covariance matrix between the features in the first output signal, and perform eigenvalue set decomposition on the covariance matrix to obtain an eigenvalue set and a corresponding eigenvector set;

[0046] The coordinate subunit is used to establish a coordinate system according to the first k largest eigenvalues ​​in the eigenvalue set and their corresponding eigenvectors in the eigenvector set to obtain a first coordinate system;

[0047] The projection subunit is used to project the first output signal into the first coordinate system to obtain the principal component characteristic signal.

[0048] As a preferred solution, the conversion unit includes an extraction subunit and a conversion subunit;

[0049] The extraction subunit is used to extract the first n principal component characteristic signals with the largest sparsity in the second output signal by frequency locking, and regard the extracted signals as the first n-order noise frequencies;

[0050] The conversion subunit is used to perform analog-to-digital conversion on the first n-order noise frequencies to obtain the digital signal.

[0051] As a preferred solution, the noise reduction module includes a sound wave unit and a feedback unit;

[0052] The sound wave unit is used to generate a sound wave signal having a signal peak value equal to that of the digital signal and a phase opposite thereto, and to perform digital-to-analog conversion on the sound wave signal to obtain an analog noise reduction signal;

[0053] The feedback unit is used to reduce the noise reduction error of the analog noise reduction signal according to the weight memory feedback and adaptive active noise reduction system to obtain the active noise reduction signal; wherein the weight of the weight memory feedback and adaptive active noise reduction system is updated according to the analog noise reduction signal and the vehicle body vibration signal, and the noise reduction error is calculated according to the in-vehicle noise signal and the analog noise reduction signal.

[0054] As a preferred solution, the weights of the weight memory feedback and adaptive active noise reduction system are updated according to the simulated noise reduction signal and the vehicle body vibration signal, specifically:

[0055] If the noise reduction error of the analog noise reduction signal is greater than a first preset value, a first weight is calculated according to the first formula, and the weight memory feedback and the weight in the adaptive active noise reduction system are updated to the first weight;

[0056] Among them, the first formula is:

[0057] F=sigmoid{(w(n)*x(n)+b)*tanh(w(n)*x(n)+b)}

[0058] Among them, w(n) is the input signal weight at the current moment, x(n) is the vehicle body vibration signal, b is a constant, and sigmoid and tanh are activation functions.

[0059] The present application also provides a storage medium, on which a computer program is stored. The computer program is called and executed by a computer to implement the multi-source noise active noise reduction method under complex working conditions as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a flow chart of a multi-source noise active noise reduction method under complex working conditions provided by an embodiment of the present application;

[0061] Figure 2 It is a diagram of weight memory and selection method provided by an embodiment of the present application;

[0062] Figure 3 is a structural diagram of an active noise reduction system provided in an embodiment of the present application;

[0063] Figure 4 is a flow chart of the active noise reduction solution provided in an embodiment of the present application;

[0064] Figure 5It is a structural schematic diagram of a multi-source noise active noise reduction device under complex working conditions provided by an embodiment of the present application. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0066] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "several" means two or more.

[0067] The embodiment of the present application provides a multi-source noise active noise reduction method under complex working conditions, which is mainly used to overcome the poor adaptability and signal lag of the active noise reduction method in the vehicle when the vehicle working conditions are complex and the excitation source input is constantly changing.

[0068] Embodiment 1:

[0069] See also Figure 1 The embodiment of the present application provides a multi-source noise active noise reduction method under complex working conditions, including S1 to S3, and the specific implementation steps are as follows:

[0070] S1. Collect the interior noise signal of the target vehicle.

[0071] Step S1 of the embodiment of the present application is specifically as follows:

[0072] The microphone is arranged at a position close to the human ear in the target vehicle, and the interior noise signal of the target vehicle is collected through the arranged microphone.

[0073] S2. Frequency lock the main component characteristic signal with the largest sparsity in the in-vehicle noise signal to obtain a digital signal.

[0074] Step S2 of the embodiment of the present application includes S2.1 to S2.2, specifically:

[0075] S2.1. Adaptively filter the in-vehicle noise signal using a filter to obtain a first output signal; wherein a weight parameter of the filter is updated according to the secondary noise signal input at the last moment;

[0076] The first output signal is standardized to ensure that each feature has the same scale; a covariance matrix between the features in the first output signal after the standardized processing is calculated, and an eigenvalue set decomposition is performed on the covariance matrix to obtain an eigenvalue set and a corresponding eigenvector set;

[0077] A linearly independent coordinate system is established according to the first k largest eigenvalues ​​in the eigenvalue set and their corresponding eigenvectors in the eigenvector set to obtain a first coordinate system;

[0078] The first output signal is projected into the first coordinate system to obtain a principal component characteristic signal.

[0079] The present embodiment S2.1 can capture the correlation between features by calculating the covariance matrix between each feature in the output signal. Since the eigenvalues ​​and eigenvectors are calculated based on the covariance matrix, and the covariance matrix is ​​a stable statistic, the constructed coordinate system also has good stability, which is helpful for subsequent signal processing and data analysis. The output signal is projected into the first coordinate system, and the principal component characteristic signal can be obtained. This method realizes the dimensionality reduction of the signal, converts the high-dimensional data into a low-dimensional representation, and retains the most important information.

[0080] S2.2, if the error value between the principal component characteristic signal and the first output signal does not meet the preset error requirement, the parameters of the sparse processing are readjusted, and the principal component characteristic signal is iteratively processed until the iterative processing result meets the error requirement, thereby obtaining a second output signal; wherein the error value may be selected from mean square error (MSE), root mean square error (RMSE) or other custom error indicators;

[0081] The first six principal component characteristic signals x(n) with the largest sparsity in the second output signal are extracted by frequency locking, and the extracted signals are regarded as the first six noise frequencies; where x(n)≈AGA -1 , A represents the orthogonal matrix after singular value decomposition of x(n), G is a diagonal matrix, and G takes the first 6 largest eigenvalues ​​in the second output signal;

[0082] The first 6 noise frequencies are converted into analog-to-digital (A / D) signals to obtain digital signals.

[0083] In this embodiment S2.2, if the error value between the main component characteristic signal and the first output signal does not meet the preset error requirement, the parameters of the sparse processing are readjusted and iterative processing is performed, so that the optimal solution can be gradually approached, the iteration efficiency is improved, and the validity of the output signal is ensured;

[0084] Moreover, by using the frequency locking technology, the main component characteristic signal in the second output signal can be accurately locked and extracted, which can accurately identify and separate the key components in the signal. The extracted main component characteristic signal is regarded as the first 6-order noise frequency, which helps to identify and distinguish the noise component and useful information in the signal, thereby ensuring that the subsequent converted signal is clear and stable.

[0085] S3: Generate an active noise reduction signal with a signal peak value equal to that of the digital signal and an opposite phase, and output the active noise reduction signal to the target vehicle.

[0086] Step S3 of the embodiment of the present application is specifically as follows:

[0087] Generate an acoustic wave signal with a signal peak value equal to that of the digital signal and an opposite phase, and perform digital-to-analog conversion (D / A conversion) on the acoustic wave signal to obtain an analog noise reduction signal;

[0088] According to the weight memory feedback and the adaptive active noise reduction system, the noise reduction error (noise reduction effect error) of the analog noise reduction signal is reduced to obtain the active noise reduction signal y(n); wherein the noise reduction error is calculated based on the in-vehicle noise signal and the analog noise reduction signal.

[0089] The active noise reduction signal y(n) is output to the target vehicle to achieve noise reduction.

[0090] Among them, the weights of the weight memory feedback and adaptive active noise reduction system are updated according to the simulated noise reduction signal and the body vibration signal, specifically:

[0091] If the noise reduction error of the simulated noise reduction signal is less than 40%, the corresponding gain coefficient in the weight memory feedback and adaptive active noise reduction system is discarded or no longer adjusted according to the discard function calculation formula;

[0092] If the noise reduction error of the simulated noise reduction signal is greater than 60%, the first weight σi is calculated according to the first formula, and the first weight σi is memorized, and the weight in the weight memory feedback and adaptive active noise reduction system is updated to the first weight σi.

[0093] Among them, the calculation formula for discarding function is:

[0094] Y=sigmoid{w(n)*x(n)+w(n)*h(n-1)+b)}

[0095] The first formula is:

[0096] F=sigmoid{(w(n)*x(n)+b)*tanh(w(n)*x(n)+b)}

[0097] In order to meet the needs of active noise reduction under complex road conditions, the weight memory feedback and the weight selection of the adaptive active noise reduction system need to have a certain memory function and keep updating in real time; the analog noise reduction signal is processed according to the weight memory feedback and the adaptive active noise reduction system to obtain the active noise reduction signal y(n), and its calculation expression is as follows:

[0098] y(n)=σi * x(n)+e(n);

[0099]

[0100] c(n)=f(n)*h(n-1)+i*g(n);

[0101] Among them, w(n) is the input signal weight at the current moment, x(n) is the vehicle body vibration signal, b is a constant, sigmoid and tanh are activation functions, f(n) represents the weight to be forgotten at time n (or step); i represents the input weight value, which is a constant between 0 and 1;

[0102] represents the associated outer product, y(n) represents the weighted memory feedback and the feedback sound wave of the adaptive active noise reduction system, e(n) is the error signal, σi represents the weight of the current memory feedback, h(n-1) represents the update of the hidden state in the previous cycle, c(n) represents the weight selection function, and g(n) is the input.

[0103] To apply the embodiments of this application, please refer to Figure 2 and Figure 3 ;

[0104] Figure 2 It is a diagram of the weight memory and selection method provided in the embodiment of the present application, which represents the weight memory feedback and weight update process of the adaptive active noise reduction system.

[0105] Figure 3 is a structural diagram of an active noise reduction system provided in an embodiment of the present application, showing the structure of a weight memory feedback and adaptive active noise reduction system; Figure 3 In the above equation, X(n) is the reference signal, which is the vibration signal of the vehicle body collected by the acceleration sensor; d(n) is the interior noise signal at the error microphone; P(z) is the primary path, which means that X(n) is the transfer function of d(n) formed by the structure and air radiation to the error microphone; W(n) represents the weight corresponding to each filter [w 1 (z),w 2 (z),…] T y(n) is the output signal, which is the secondary sound wave signal emitted by the speaker driven by the controller; S(z) is the secondary path, the transfer function of the sound wave from the secondary speaker to the error microphone; ys (n) is the speaker output signal collected at the error speaker; e(n) is the error signal, which represents the difference between d(n) and y s (n) The residual error after cancellation, i.e., the actual collected signal of the error microphone;

[0106] Compared with the FxLMS system (filtered minimum mean square error system), the weighted memory feedback and adaptive active noise reduction system has feedback iteration and memory functions. After the principal component analysis (PCA) is completed, the sparse signal can reduce a lot of computational difficulty to meet the real-time requirements of the weighted memory feedback and adaptive active noise reduction system. Among them, the working principle of the weighted memory is to capture and keep the gains with better memory effects in the weighted gains of the entire time series according to the noise reduction effect of the actual output signal, and discard the gain weights with poor effects, so as to enhance the adaptive ability of the entire active noise reduction scheme and improve the robust control effect.

[0107] To apply the embodiments of this application, please refer to Figure 4 , Figure 4 It is a flow chart of the active noise reduction solution provided in an embodiment of the present application, which shows the operation process of performing in-vehicle noise reduction in the first embodiment of the present application; wherein w(n-1) represents the weighted value under the filter, which is related to the frequency of the initial input signal; σi(n-1) represents the weight selected by the noise reduction network for the input signal in the previous stage according to the noise reduction effect; y(n-1) represents the output result of the entire network in the previous cycle stage.

[0108] In this embodiment S3, by generating a sound wave signal with a signal peak value equal to that of the digital signal and a phase opposite thereto, a noise reduction signal that counteracts the noise can be accurately generated. This method is based on the physical properties of sound waves and ensures that the noise reduction signal and the noise are completely opposite in phase, thereby achieving effective noise cancellation.

[0109] Moreover, in practical applications, the noise environment is often complex and changeable. By dynamically adjusting the weights according to the noise reduction error, the current noise characteristics can be optimized, thereby improving the accuracy and effect of noise reduction. The weight memory feedback and adaptive active noise reduction system can selectively update and maintain memory, update the system input and output states, and thus improve the system's computational accuracy and robustness.

[0110] Overall, this embodiment has the following beneficial effects:

[0111] This application performs frequency locking on the main component characteristic signal with the greatest sparsity in the in-car noise, which means that the most important and representative component of the in-car noise can be accurately identified. The digital signal obtained by frequency locking is an important basis for reflecting the in-car noise characteristics in real time; therefore, based on this digital signal, a corresponding active noise reduction signal can be quickly generated. The signal peak of this noise reduction signal is equal to the signal peak of the digital signal, but the phase is opposite, thereby ensuring the maximization of the noise reduction effect. In addition, the frequency locking technology can track the changes in the in-car noise in real time. When the noise characteristics change, the frequency locking parameters can be quickly adjusted to ensure that the noise reduction signal always maintains the best phase and amplitude match with the noise signal. This real-time adjustment capability significantly enhances the adaptability of this solution, enabling it to maintain a stable noise reduction effect in different noise environments;

[0112] In addition, the input noise signal is subjected to principal component analysis and sparse processing to extract the main characteristic peaks, which reduces the computational memory and solves the problem of secondary channel signal lag, thus reducing the hardware cost while ensuring the noise reduction performance. The weight memory and selection function can be used to learn and feedback the working conditions, selectively update and save the memory, solve the problem of poor adaptability of this solution, and improve the robustness of the system.

[0113] Embodiment 2:

[0114] See also Figure 5 , the embodiment of the present application provides a multi-source noise active noise reduction device under complex working conditions, including a collection module 10, a locking module 20 and a noise reduction module 30;

[0115] Wherein, the acquisition module 10 is used to collect the interior noise signal of the target vehicle;

[0116] A locking module 20, used for frequency locking the main component characteristic signal with the largest sparsity in the in-vehicle noise signal to obtain a digital signal;

[0117] The noise reduction module 30 is used to generate an active noise reduction signal having a signal peak value equal to that of the digital signal and an opposite phase, and output the active noise reduction signal to the target vehicle.

[0118] In one embodiment, the acquisition module 10 is specifically:

[0119] The microphone is arranged at a position close to the human ear in the target vehicle, and the interior noise signal of the target vehicle is collected through the arranged microphone.

[0120] In one embodiment, the locking module 20 includes a filtering unit, a decomposition subunit, a coordinate subunit, a projection subunit, an error unit, an extraction subunit and a conversion subunit, specifically:

[0121] The filtering unit is used to adaptively filter the in-vehicle noise signal using a filter to obtain a first output signal; wherein the weight parameter of the filter is updated according to the secondary noise signal input at the last moment;

[0122] The decomposition subunit is used to perform standardization processing on the first output signal to ensure that each feature has the same scale; calculate the covariance matrix between the features in the first output signal after the standardization processing, and perform eigenvalue set decomposition on the covariance matrix to obtain an eigenvalue set and a corresponding eigenvector set;

[0123] A coordinate subunit is used to establish a linearly independent coordinate system according to the first k largest eigenvalues ​​in the eigenvalue set and their corresponding eigenvectors in the eigenvector set to obtain a first coordinate system;

[0124] The projection subunit is used to project the first output signal into the first coordinate system to obtain a principal component characteristic signal.

[0125] The filter unit, decomposition subunit, coordinate subunit and projection subunit of this embodiment can capture the correlation between features by calculating the covariance matrix between each feature in the output signal. Since the eigenvalues ​​and eigenvectors are calculated based on the covariance matrix, and the covariance matrix is ​​a stable statistic, the constructed coordinate system also has good stability, which is helpful for subsequent signal processing and data analysis. The output signal is projected into the first coordinate system to obtain the principal component characteristic signal. This method realizes the dimensionality reduction of the signal, converts the high-dimensional data into a low-dimensional representation, and retains the most important information.

[0126] An error unit, used for re-adjusting the parameters of the sparse processing if the error value between the principal component characteristic signal and the first output signal does not meet the preset error requirement, and iteratively processing the principal component characteristic signal until the iterative processing result meets the error requirement, thereby obtaining a second output signal; wherein the error value may be selected from mean square error (MSE), root mean square error (RMSE) or other custom error indicators;

[0127] The first six principal component characteristic signals x(n) with the largest sparsity in the second output signal are extracted by frequency locking, and the extracted signals are regarded as the first six noise frequencies; where x(n)≈AGA -1 , A represents the orthogonal matrix after singular value decomposition of x(n), G is a diagonal matrix, and G takes the first 6 largest eigenvalues ​​in the second output signal;

[0128] The conversion subunit is used to perform analog-to-digital conversion (A / D conversion) on the first 6-order noise frequencies to obtain digital signals.

[0129] In the error unit, the extraction subunit and the conversion subunit of this embodiment, if the error value between the main component feature signal and the first output signal does not meet the preset error requirement, the parameters of the sparse processing are readjusted and iterative processing is performed, so that the optimal solution can be gradually approached, the iteration efficiency is improved, and the validity of the output signal is guaranteed;

[0130] Moreover, by using the frequency locking technology, the main component characteristic signal in the second output signal can be accurately locked and extracted, which can accurately identify and separate the key components in the signal. The extracted main component characteristic signal is regarded as the first 6-order noise frequency, which helps to identify and distinguish the noise component and useful information in the signal, thereby ensuring that the subsequent converted signal is clear and stable.

[0131] In one embodiment, the noise reduction module 30 includes a sound wave unit, a feedback unit, and an actual noise reduction unit;

[0132] The sound wave unit is used to generate a sound wave signal with a signal peak value equal to that of the digital signal and a phase opposite thereto, and to perform digital-to-analog conversion (D / A conversion) on the sound wave signal to obtain an analog noise reduction signal;

[0133] The feedback unit is used to reduce the noise reduction error (noise reduction effect error) of the analog noise reduction signal according to the weight memory feedback and the adaptive active noise reduction system to obtain the active noise reduction signal y(n); wherein the noise reduction error is calculated based on the in-vehicle noise signal and the analog noise reduction signal.

[0134] The actual noise reduction unit is also used to output the active noise reduction signal y(n) to the target vehicle to achieve noise reduction.

[0135] Among them, the weights of the weight memory feedback and adaptive active noise reduction system are updated according to the simulated noise reduction signal and the body vibration signal, specifically:

[0136] If the noise reduction error of the simulated noise reduction signal is less than 40%, the corresponding gain coefficient in the weight memory feedback and adaptive active noise reduction system is discarded or no longer adjusted according to the discard function calculation formula;

[0137] If the noise reduction error of the simulated noise reduction signal is greater than 60%, the first weight σi is calculated according to the first formula, and the first weight σi is memorized, and the weight in the weight memory feedback and adaptive active noise reduction system is updated to the first weight σi.

[0138] Among them, the calculation formula for discarding function is:

[0139] Y=sigmoid{w(n)*x(n)+w(n)*h(n-1)+b)}

[0140] The first formula is:

[0141] F=sigmoid{(w(n)*x(n)+b)*tanh(w(n)*x(n)+b)}

[0142] In order to meet the needs of active noise reduction under complex road conditions, the weight memory feedback and the weight selection of the adaptive active noise reduction system need to have a certain memory function and keep updating in real time; the analog noise reduction signal is processed according to the weight memory feedback and the adaptive active noise reduction system to obtain the active noise reduction signal y(n), and its calculation expression is as follows:

[0143] y(n)=σi * x(n)+e(n);

[0144]

[0145] c(n)=f(n)*h(n-1)+i*g(n);

[0146] Among them, w(n) is the input signal weight at the current moment, x(n) is the vehicle body vibration signal, b is a constant, sigmoid and tanh are activation functions, f(n) represents the weight to be forgotten at time n (or step); i represents the input weight value, which is a constant between 0 and 1;

[0147] represents the associated outer product, y(n) represents the weighted memory feedback and the feedback sound wave of the adaptive active noise reduction system, e(n) is the error signal, w(n-1) represents the weight of the previous cycle, σi(n-1) represents the weight under the previous working condition, σi represents the weight of the current memory feedback, h(n-1) represents the update of the hidden state under the previous cycle, c(n) represents the weight selection function, and g(n) is the input.

[0148] To apply the embodiments of this application, please refer to Figure 2 and Figure 3 ;

[0149] Figure 2 It is a diagram of the weight memory and selection method provided in the embodiment of the present application, which represents the weight memory feedback and weight update process of the adaptive active noise reduction system.

[0150] Figure 3 is a structural diagram of an active noise reduction system provided in an embodiment of the present application, showing the structure of a weight memory feedback and adaptive active noise reduction system; Figure 3 In the above equation, X(n) is the reference signal, which is the vibration signal of the vehicle body collected by the acceleration sensor; d(n) is the interior noise signal at the error microphone; P(z) is the primary path, which means that X(n) is the transfer function of d(n) formed by the structure and air radiation to the error microphone; W(n) represents the weight corresponding to each filter [w 1(z),w 2 (z),…] T y(n) is the output signal, which is the secondary sound wave signal emitted by the speaker driven by the controller; S(z) is the secondary path, the transfer function of the sound wave from the secondary speaker to the error microphone; y s (n) is the speaker output signal collected at the error speaker; e(n) is the error signal, which represents the difference between d(n) and y s (n) The residual error after cancellation, i.e., the actual collected signal of the error microphone;

[0151] Compared with the FxLMS system (filtered minimum mean square error system), the weighted memory feedback and adaptive active noise reduction system has feedback iteration and memory functions. After the principal component analysis (PCA) is completed, the sparse signal can reduce a lot of computational difficulty to meet the real-time requirements of the weighted memory feedback and adaptive active noise reduction system. Among them, the working principle of the weighted memory is to capture and keep the gains with better memory effects in the weighted gains of the entire time series according to the noise reduction effect of the actual output signal, and discard the gain weights with poor effects, so as to enhance the adaptive ability of the entire active noise reduction scheme and improve the robust control effect.

[0152] To apply the embodiments of this application, please refer to Figure 4 , Figure 4 It is a flow chart of the active noise reduction solution provided in an embodiment of the present application, which shows the operation process of performing in-vehicle noise reduction in the first embodiment of the present application; wherein w(n-1) represents the weighted value under the filter, which is related to the frequency of the initial input signal; σi(n-1) represents the weight selected by the noise reduction network for the input signal in the previous stage according to the noise reduction effect; y(n-1) represents the output result of the entire network in the previous cycle stage.

[0153] The noise reduction module 30 of this embodiment can accurately generate a noise reduction signal that counteracts the noise by generating a sound wave signal that is equal to the signal peak value of the digital signal and opposite in phase. This method is based on the physical properties of sound waves and ensures that the noise reduction signal and the noise are completely opposite in phase, thereby achieving effective noise cancellation.

[0154] Moreover, in practical applications, the noise environment is often complex and changeable. By dynamically adjusting the weights according to the noise reduction error, the current noise characteristics can be optimized, thereby improving the accuracy and effect of noise reduction. The weight memory feedback and adaptive active noise reduction system can selectively update and maintain memory, update the system input and output states, and thus improve the system's computational accuracy and robustness.

[0155] Overall, this embodiment has the following beneficial effects:

[0156] This application performs frequency locking on the main component characteristic signal with the greatest sparsity in the in-car noise, which means that the most important and representative component of the in-car noise can be accurately identified. The digital signal obtained by frequency locking is an important basis for reflecting the in-car noise characteristics in real time; therefore, based on this digital signal, a corresponding active noise reduction signal can be quickly generated. The signal peak of this noise reduction signal is equal to the signal peak of the digital signal, but the phase is opposite, thereby ensuring the maximization of the noise reduction effect. In addition, the frequency locking technology can track the changes in the in-car noise in real time. When the noise characteristics change, the frequency locking parameters can be quickly adjusted to ensure that the noise reduction signal always maintains the best phase and amplitude match with the noise signal. This real-time adjustment capability significantly enhances the adaptability of this solution, enabling it to maintain a stable noise reduction effect in different noise environments;

[0157] In addition, the input noise signal is subjected to principal component analysis and sparse processing to extract the main characteristic peaks, which reduces the computational memory and solves the problem of secondary channel signal lag, thus reducing the hardware cost while ensuring the noise reduction performance. The weight memory and selection function can be used to learn and feedback the working conditions, selectively update and save the memory, solve the problem of poor adaptability of this solution, and improve the robustness of the system.

[0158] Embodiment three:

[0159] An embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to perform the multi-source noise active noise reduction method under a complex working condition;

[0160] Wherein, the multi-source noise active noise reduction method under complex working conditions, if implemented in the form of a software functional unit and used as an independent product, can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0161] The above are preferred embodiments of the present invention. It should be noted that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A multi-source noise active noise reduction method under complex working conditions, characterized in that: include: Collect the interior noise signal of the target vehicle; Frequency locking is performed on the main component characteristic signal with the largest sparsity in the in-vehicle noise signal to obtain a digital signal; An active noise reduction signal having a signal peak value equal to that of the digital signal and an opposite phase is generated, and the active noise reduction signal is output to the target vehicle.

2. The multi-source noise active noise reduction method under complex working conditions as claimed in claim 1, characterized in that: The frequency of the main component characteristic signal with the largest sparsity in the in-vehicle noise signal is locked to obtain a digital signal, which is specifically: Adaptively filtering the in-vehicle noise signal to obtain a first output signal; Performing sparse processing on the first output signal to obtain the principal component feature signal with the maximum sparsity; If the error value between the principal component characteristic signal and the first output signal does not meet the preset error requirement, the parameters of the sparse processing are readjusted, and the principal component characteristic signal is iteratively processed until the iterative processing result meets the error requirement, thereby obtaining a second output signal; Perform analog-to-digital conversion on the first n principal component feature signals with the largest sparsity in the second output signal to obtain the digital signal.

3. The multi-source noise active noise reduction method under complex working conditions as claimed in claim 2, characterized in that: The first output signal is subjected to sparse processing to obtain the principal component characteristic signal with the largest sparsity, specifically: Calculating a covariance matrix between features in the first output signal, and performing eigenvalue set decomposition on the covariance matrix to obtain an eigenvalue set and a corresponding eigenvector set; Establishing a coordinate system according to the first k largest eigenvalues ​​in the eigenvalue set and their corresponding eigenvectors in the eigenvector set to obtain a first coordinate system; The first output signal is projected into the first coordinate system to obtain the principal component characteristic signal.

4. The multi-source noise active noise reduction method under complex working conditions as claimed in claim 2, characterized in that: Performing analog-to-digital conversion on the first n principal component feature signals with the largest sparsity in the second output signal to obtain the digital signal is specifically: Extracting the first n principal component characteristic signals with the largest sparsity in the second output signal by frequency locking, and treating the extracted signals as the first n-order noise frequencies; Perform analog-to-digital conversion on the first n-order noise frequencies to obtain the digital signal.

5. The multi-source noise active noise reduction method under complex working conditions as claimed in claim 1, characterized in that: Generate an active noise reduction signal with a signal peak value equal to that of the digital signal and an opposite phase, specifically: Generating a sound wave signal having a signal peak value equal to that of the digital signal and an opposite phase, and performing digital-to-analog conversion on the sound wave signal to obtain an analog noise reduction signal; The active noise reduction signal is obtained by reducing the noise reduction error of the analog noise reduction signal according to the weight memory feedback and adaptive active noise reduction system; wherein the weight of the weight memory feedback and adaptive active noise reduction system is updated according to the analog noise reduction signal and the vehicle body vibration signal, and the noise reduction error is calculated according to the in-vehicle noise signal and the analog noise reduction signal.

6. The multi-source noise active noise reduction method under complex working conditions as claimed in claim 5, characterized in that: The weights of the weight memory feedback and adaptive active noise reduction system are updated according to the simulated noise reduction signal and the vehicle body vibration signal, specifically: If the noise reduction error of the analog noise reduction signal is greater than a first preset value, a first weight is calculated according to the first formula, and the weight memory feedback and the weight in the adaptive active noise reduction system are updated to the first weight; Among them, the first formula is: F=sigmoid{(w(n)*x(n)+b)*tanh(w(n)*x(n)+b)} Among them, w(n) is the input signal weight at the current moment, x(n) is the vehicle body vibration signal, b is a constant, and sigmoid and tanh are activation functions.

7. A multi-source noise active noise reduction device under complex working conditions, characterized in that: Including acquisition module, locking module and noise reduction module; Wherein, the acquisition module is used to collect the interior noise signal of the target vehicle; The locking module is used to perform frequency locking on the main component characteristic signal with the largest sparsity in the in-vehicle noise signal to obtain a digital signal; The noise reduction module is used to generate an active noise reduction signal having a signal peak value equal to that of the digital signal and an opposite phase, and output the active noise reduction signal to the target vehicle.

8. The multi-source noise active noise reduction device under complex working conditions as claimed in claim 7, characterized in that: The locking module includes a filtering unit, a sparse unit, an error unit and a conversion unit; The filtering unit is used to adaptively filter the in-vehicle noise signal to obtain a first output signal; The sparseness unit is used to perform sparseness processing on the first output signal to obtain the principal component feature signal with the maximum sparsity; The error unit is used to readjust the parameters of the sparse processing if the error value between the principal component feature signal and the first output signal does not meet the preset error requirement, and iteratively process the principal component feature signal until the iterative processing result meets the error requirement to obtain the second output signal; The conversion unit is used to perform analog-to-digital conversion on the first n principal component feature signals with the largest sparsity in the second output signal to obtain the digital signal.

9. The multi-source noise active noise reduction device under complex working conditions as claimed in claim 8, characterized in that: The sparseness unit includes a decomposition subunit, a coordinate subunit and a projection subunit; The decomposition subunit is used to calculate the covariance matrix between the features in the first output signal, and perform eigenvalue set decomposition on the covariance matrix to obtain an eigenvalue set and a corresponding eigenvector set; The coordinate subunit is used to establish a coordinate system according to the first k largest eigenvalues ​​in the eigenvalue set and their corresponding eigenvectors in the eigenvector set to obtain a first coordinate system; The projection subunit is used to project the first output signal into the first coordinate system to obtain the principal component characteristic signal.

10. The multi-source noise active noise reduction device under complex working conditions as claimed in claim 8, characterized in that: The conversion unit includes an extraction subunit and a conversion subunit; The extraction subunit is used to extract the first n principal component characteristic signals with the largest sparsity in the second output signal by frequency locking, and regard the extracted signals as the first n-order noise frequencies; The conversion subunit is used to perform analog-to-digital conversion on the first n-order noise frequencies to obtain the digital signal.

11. The multi-source noise active noise reduction device under complex working conditions as claimed in claim 7, characterized in that: The noise reduction module includes a sound wave unit and a feedback unit; The sound wave unit is used to generate a sound wave signal having a signal peak value equal to that of the digital signal and a phase opposite thereto, and to perform digital-to-analog conversion on the sound wave signal to obtain an analog noise reduction signal; The feedback unit is used to reduce the noise reduction error of the analog noise reduction signal according to the weight memory feedback and adaptive active noise reduction system to obtain the active noise reduction signal; wherein the weight of the weight memory feedback and adaptive active noise reduction system is updated according to the analog noise reduction signal and the vehicle body vibration signal, and the noise reduction error is calculated according to the in-vehicle noise signal and the analog noise reduction signal.

12. The multi-source noise active noise reduction device under complex working conditions according to claim 11, characterized in that: The weights of the weight memory feedback and adaptive active noise reduction system are updated according to the simulated noise reduction signal and the vehicle body vibration signal, specifically: If the noise reduction error of the analog noise reduction signal is greater than a first preset value, a first weight is calculated according to the first formula, and the weight memory feedback and the weight in the adaptive active noise reduction system are updated to the first weight; Among them, the first formula is: F=sigmoid{(w(n)*x(n)+b)*tanh(w(n)*x(n)+b)} Among them, w(n) is the input signal weight at the current moment, x(n) is the vehicle body vibration signal, b is a constant, and sigmoid and tanh are activation functions.

13. A storage medium, characterized in that: The storage medium stores a computer program, which is called and executed by a computer to implement a multi-source noise active noise reduction method under complex working conditions as described in any one of claims 1 to 6 above.

Images