In-vehicle noise reduction method, device, equipment, storage medium and product

Abstract

This application relates to the field of automotive noise reduction technology, and more particularly to an in-vehicle noise reduction method, device, equipment, storage medium, and product. The method is based on an adaptive filtering algorithm to generate multiple control filter weight coefficients; based on a preset feature extraction algorithm, it determines the mapping relationship between preset training reference signals and preset training error signals and the control filter weight coefficients; based on the mapping relationship, it extracts the control filter weight coefficients corresponding to the current reference signal and the current error signal, and uses them as noise reduction coefficients; it convolves the noise reduction coefficients with the current reference signal to generate the target noise-reduced signal. This method, by utilizing an adaptive filtering algorithm and feature extraction mapping technology, achieves rapid matching and dynamic adjustment of filter weight coefficients, effectively improving noise reduction speed and applicability, enabling the in-vehicle active noise reduction system to maintain efficient and stable noise reduction performance even in complex and variable noise environments.

Description

Technical Field

[0001] This application relates to the field of automotive noise reduction technology, and in particular to a method, device, equipment, storage medium and product for in-vehicle noise reduction. Background Technology

[0002] With the development of the automotive industry, in-vehicle noise has become increasingly important for driving and riding comfort, making its reduction a crucial factor in enhancing user experience. Traditional in-vehicle noise reduction technologies primarily rely on the application of passive sound insulation materials, such as sound-absorbing cotton and sound-absorbing pads. While these methods can reduce noise to some extent, their effectiveness is limited, and they are not particularly effective at controlling low-frequency noise. Furthermore, passive noise reduction increases vehicle weight, which is detrimental to the trend of energy-saving and environmentally friendly automotive design. Active noise control (ANC) technology has become a research hotspot in recent years due to its effective suppression of low-frequency noise. Traditional active noise control systems often use adaptive filtering algorithms to generate noise reduction signals, but these suffer from slow convergence speed and poor adaptability, making it difficult to achieve ideal noise reduction effects in complex and ever-changing in-vehicle noise environments. Therefore, improving the speed and applicability of in-vehicle noise reduction has become an urgent problem to be solved. Summary of the Invention

[0003] The main objective of this application is to provide a method, apparatus, device, storage medium, and product for in-vehicle noise reduction, aiming to solve the technical problem of how to improve the speed and applicability of in-vehicle noise reduction.

[0004] To achieve the above objectives, this application provides an in-vehicle noise reduction method, the method comprising:

[0005] Based on an adaptive filtering algorithm, multiple control filter weight coefficients are generated.

[0006] Based on a preset feature extraction algorithm, the mapping relationship between the preset training reference signal and the preset training error signal and the weight coefficients of the control filter is determined;

[0007] Based on the mapping relationship, the control filter weight coefficients corresponding to the current reference signal and the current error signal are extracted and used as noise reduction coefficients;

[0008] The noise reduction coefficients are convolved with the current reference signal to generate the target noise-reduced signal.

[0009] In one embodiment, the step of generating multiple control filter weight coefficients based on an adaptive filtering algorithm includes:

[0010] Use noise information as input to a preset filter;

[0011] Initialize the weight coefficients and length of the preset filter;

[0012] Based on the adaptive filtering algorithm, the weight coefficients of the preset filter are adjusted to obtain the weight coefficients of the plurality of control filters.

[0013] In one embodiment, the step of determining the mapping relationship between the preset training reference signal and the preset training error signal and the weight coefficients of the control filter based on a preset feature extraction algorithm includes:

[0014] The preset training reference signal and the preset training error signal are preprocessed to obtain a preprocessed signal, wherein the preprocessing includes one or more of DC component removal processing, noise removal processing and normalization processing;

[0015] Based on the preset feature extraction algorithm, feature extraction is performed on the preprocessed signal to obtain noise features;

[0016] The noise features are combined with the corresponding control filter weight coefficients to obtain the mapping relationship.

[0017] In one embodiment, after the step of combining the noise features with the corresponding control filter weight coefficients to obtain the mapping relationship, the method further includes:

[0018] Based on the noise characteristics, the weight coefficients of the multiple control filters, and the mapping relationship, a mapping model is constructed;

[0019] The training performance of the mapping model is evaluated using a loss function;

[0020] The mapping model that passes the evaluation is validated, and the hyperparameters of the mapping model are adjusted based on the validation results.

[0021] In one embodiment, the step of extracting the control filter weight coefficients corresponding to the current reference signal and the current error signal based on the mapping relationship, and using them as noise reduction coefficients, includes:

[0022] The preprocessing is performed on the current reference signal and the current error signal to obtain the current processed signal;

[0023] Based on the preset feature extraction algorithm, features are extracted from the currently processed signal to obtain the current feature matrix;

[0024] The current feature matrix is ​​input into the mapping model to obtain the corresponding control filter weight coefficients, which are then used as the noise reduction coefficients.

[0025] In one embodiment, the step of convolving the noise reduction coefficients with the current reference signal to generate the target noise-reduced signal includes:

[0026] Align the initial position of the noise reduction coefficient with the starting point of the current reference signal;

[0027] The weighted sum at each position is determined by moving the noise reduction coefficients point by point along the current reference signal;

[0028] Based on the weighted sum, a noise-reduced signal is generated;

[0029] The denoised signal is verified based on the current error signal, and the verified denoised signal is taken as the target denoised signal.

[0030] Furthermore, to achieve the above objectives, this application also proposes an in-vehicle noise reduction device, which includes:

[0031] The coefficient determination module is used to generate multiple control filter weight coefficients based on an adaptive filtering algorithm.

[0032] The mapping determination module is used to determine the mapping relationship between the preset training reference signal and the preset training error signal and the weight coefficients of the control filter based on the preset feature extraction algorithm.

[0033] The coefficient extraction module is used to extract the control filter weight coefficients corresponding to the current reference signal and the current error signal based on the mapping relationship, and use them as noise reduction coefficients.

[0034] The target module is used to convolve the noise reduction coefficients with the current reference signal to generate a target noise reduction signal.

[0035] In addition, to achieve the above objectives, this application also proposes an in-vehicle noise reduction device, the device comprising: a memory, a processor, and an in-vehicle noise reduction program stored in the memory and executable on the processor, the in-vehicle noise reduction program being configured to implement the steps of the in-vehicle noise reduction method as described above.

[0036] In addition, to achieve the above objectives, this application also proposes a storage medium storing an in-vehicle noise reduction program, which, when executed by a processor, implements the steps of the in-vehicle noise reduction method described above.

[0037] In addition, to achieve the above objectives, this application also proposes a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the in-vehicle noise reduction method described above.

[0038] This application generates multiple control filter weight coefficients based on an adaptive filtering algorithm; it determines the mapping relationship between the preset training reference signal and preset training error signal and the control filter weight coefficients based on a preset feature extraction algorithm; based on the mapping relationship, it extracts the control filter weight coefficients corresponding to the current reference signal and the current error signal, and uses them as noise reduction coefficients; it then convolves the noise reduction coefficients with the current reference signal to generate the target noise-reduced signal. This application, by utilizing an adaptive filtering algorithm and feature extraction mapping technology, achieves rapid matching and dynamic adjustment of filter weight coefficients, effectively improving noise reduction speed and applicability, enabling the in-vehicle active noise cancellation system to maintain efficient and stable noise reduction performance even in complex and variable noise environments. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the first embodiment of the in-vehicle noise reduction method of this application;

[0040] Figure 2 This is a schematic diagram of a sub-process in the second embodiment of the in-vehicle noise reduction method of this application;

[0041] Figure 3 This is a schematic diagram of a sub-process in the third embodiment of the in-vehicle noise reduction method of this application;

[0042] Figure 4 A flowchart for establishing the set of noise reduction weights and training the CNN model;

[0043] Figure 5 This is a schematic diagram showing the sub-band division of the reference signal and error signal;

[0044] Figure 6 Here is a block diagram of a 2D CNN model;

[0045] Figure 7 This is a schematic diagram of the module structure of the in-vehicle noise reduction device according to an embodiment of this application;

[0046] Figure 8 This is a schematic diagram of the device structure of the hardware operating environment involved in the in-vehicle noise reduction method in the embodiments of this application.

[0047] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0048] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this application.

[0049] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0050] It should be noted that with the development of the automotive industry, the impact of in-vehicle noise on driving comfort has become increasingly significant, making noise reduction a crucial factor in improving user experience. Traditional in-vehicle noise reduction technologies primarily rely on the application of passive sound insulation materials, such as sound-absorbing cotton and sound-absorbing pads. While these methods can reduce noise to some extent, their effectiveness is limited, and they are not very effective at controlling low-frequency noise. Furthermore, passive noise reduction methods increase vehicle weight, which is detrimental to the trend of energy-saving and environmentally friendly automotive design. Active noise control (ANC) technology has become a research hotspot in recent years due to its effective suppression of low-frequency noise. Traditional active noise control systems often use adaptive filtering algorithms to generate noise reduction signals, but these suffer from slow convergence speed and poor adaptability, making it difficult to achieve ideal noise reduction effects in complex and ever-changing in-vehicle noise environments. Therefore, improving the speed and applicability of in-vehicle noise reduction has become an urgent problem to be solved.

[0051] The main solution of this application is as follows: Based on an adaptive filtering algorithm, multiple control filter weight coefficients are generated; based on a preset feature extraction algorithm, the mapping relationship between the preset training reference signal and the preset training error signal and the control filter weight coefficients is determined; based on the mapping relationship, the control filter weight coefficients corresponding to the current reference signal and the current error signal are extracted and used as denoising coefficients; the denoising coefficients are convolved with the current reference signal to generate the target denoised signal.

[0052] This application utilizes adaptive filtering algorithms and feature extraction mapping technology to achieve rapid matching and dynamic adjustment of filter weight coefficients, effectively improving noise reduction speed and applicability, enabling the in-vehicle active noise reduction system to maintain efficient and stable noise reduction performance even in complex and variable noise environments.

[0053] It should be noted that the executing entity of the method in this embodiment can be a computing service device with data processing, network communication, and program execution functions, or it can be the aforementioned in-vehicle noise reduction device with the same or similar functions. This embodiment and the following embodiments will be described using an in-vehicle noise reduction device as an example.

[0054] Based on this, a first embodiment of the in-vehicle noise reduction method of this application is proposed. Please refer to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the in-vehicle noise reduction method of this application.

[0055] In this embodiment, the method includes the following steps:

[0056] S1: Based on the adaptive filtering algorithm, generate multiple control filter weight coefficients;

[0057] It's important to note that adaptive filtering algorithms dynamically adjust filter weights to minimize error signals. They update filter parameters in real-time based on changes in the input signal, enabling the system to adapt to different environments and maintain optimal filtering performance. Common adaptive filtering algorithms include Normalized Least Mean Square (NLMS) and Recursive Least Squares (RLS), which adjust filter weights using different update rules to ensure good tracking capability against changing noise environments. Controlling filter weights is a crucial parameter, controlling the output signal shape. The magnitude and arrangement of the weights directly determine how the filter processes the input signal. By adjusting the weights, the filter can optimize the processing of the reference signal to generate a noise-reduced signal suitable for the current noise environment. Multiple sets of controlled filter weights can address different noise patterns, improving the system's adaptability and response speed.

[0058] Specifically, initial reference and error signals are first acquired from in-vehicle sensors and input to an adaptive filter. The filter's initial weight coefficients are set to zero or a small random value to ensure that the system does not excessively interfere with the input signal in the initial state. Adaptive filtering algorithms, such as NLMS or RLS, use these initial signals to begin adjusting the filter's weight coefficients. The length of the filter (i.e., the number of weight coefficients) is determined by the system's design requirements, typically depending on the noise complexity and the system's computational capabilities.

[0059] Furthermore, the adaptive filter iteratively processes each set of input reference and error signals, adjusting the current filter weights based on the magnitude and direction of the error signal. Through gradual adjustments, the filter outputs a set of updated weights after each iteration, which better match the current noise characteristics. The system continuously performs these adjustments, generating multiple sets of control filter weights, each representing the optimal noise reduction strategy under different noise scenarios. These multiple sets of weights can be stored for use in different noise environments or for subsequent weight mapping and optimization processes.

[0060] By generating multiple control filter weight coefficients using an adaptive filtering algorithm, the system can pre-prepare a set of weight coefficients suitable for various noise environments. In actual operation, the system can quickly select the weight coefficients most suitable for the current noise characteristics from these sets, avoiding the complexity and latency of real-time calculations and significantly improving the system's response speed. Different noise environments have different spectral characteristics, and the adaptive filtering algorithm can generate multiple sets of weight coefficients to match various situations based on changes in the noise environment. This dynamic adjustment capability allows the system to flexibly cope with diverse noise sources inside the vehicle, such as road bumps, engine noise, and ambient noise, significantly improving the applicability and stability of the in-vehicle noise reduction system. Through repeated adjustments and optimizations, the generated control filter weight coefficients can minimize error signals, ensuring the system is in an optimal noise reduction state at every moment. This not only improves the noise cancellation effect but also makes the in-vehicle environment quieter, improving the riding experience and comfort. This step provides the foundation for the subsequent rapid matching of filter weight coefficients with noise characteristics, greatly improving the overall performance of the in-vehicle noise reduction system.

[0061] S2: Based on a preset feature extraction algorithm, determine the mapping relationship between the preset training reference signal and the preset training error signal and the weight coefficients of the control filter;

[0062] It's important to note that preset feature extraction algorithms refer to techniques for analyzing and extracting features from signals, transforming time-domain or frequency-domain signals into data formats that reflect their essential characteristics. Commonly used feature extraction algorithms include Fourier Transform (FFT), Short-Time Fourier Transform (STFT), and Mel Spectrum Analysis. These algorithms can transform complex signals into easily analyzed and processed feature parameters, providing a foundation for subsequent signal processing. Mapping relationships refer to establishing the correlation between input signal features and output control parameters (such as filter weight coefficients) through models or algorithms. This relationship reflects how signal features influence or determine the filter's adjustment method and is crucial for achieving intelligent noise reduction. Mapping relationships can be trained and optimized using data-driven methods such as machine learning and deep learning.

[0063] Specifically, firstly, a large number of pre-defined training reference signals and error signals are collected from multiple noisy environments. These signals can be data collected during actual driving or data generated in a laboratory simulation environment. Next, these signals are processed using a pre-defined feature extraction algorithm, converting them into feature matrices, such as spectral features, energy distribution, and phase information. The extracted features accurately describe the time-frequency characteristics of the signal. These features, along with the corresponding control filter weights, form the training dataset.

[0064] Furthermore, by selecting appropriate machine learning or deep learning models (such as Convolutional Neural Networks (CNNs) and Support Vector Machines (SVMs), the extracted signal features are correlated with the control filter weights during training. The model learns the complex relationship between input features and output weights through extensive training data and continuously adjusts its parameters to optimize prediction accuracy. After training, the model can accurately predict the optimal control filter weights corresponding to the current signal features, achieving real-time and rapid mapping and ensuring that the system can automatically adjust to its optimal state in different noise environments.

[0065] By establishing feature extraction algorithms and mapping relationships, the system can quickly identify the characteristics of the current noise environment and accurately match the most suitable control filter weight coefficients. Compared to traditional successive optimization methods, the use of mapping relationships significantly shortens the selection time of weight coefficients, thereby improving the response speed of the noise reduction system. The mapping relationship allows the system to automatically select the optimal filter weight coefficients based on the current environment, without human intervention or manual adjustment. This intelligent adjustment capability enables the system to flexibly cope with different noise scenarios. Whether it's high-frequency or low-frequency noise, the system can make the best response in real time, improving the effectiveness and stability of noise reduction. Through precise mapping relationships, the system can always maintain the optimal weight coefficient working state, ensuring that the output noise-reduced signal can cancel out the actual noise to the greatest extent. This optimization can significantly reduce error signals inside the vehicle, making the in-vehicle environment quieter and more comfortable, thereby improving the passenger experience and the overall quality of the vehicle. This step, by establishing an intelligent correlation between signal features and control parameters, achieves adaptive optimization of the noise reduction system, significantly improving the efficiency and effectiveness of in-vehicle noise control.

[0066] S3: Based on the mapping relationship, extract the control filter weight coefficients corresponding to the current reference signal and the current error signal, and use them as noise reduction coefficients;

[0067] It should be noted that the current reference signal is the latest noise signal collected by the in-vehicle active noise cancellation system during real-time operation, typically captured in real-time by multiple sensors (such as microphones) within the vehicle. This signal reflects the characteristics of current noise sources within the vehicle, including noise from the engine, tires, road surface, etc. The current error signal is the actual noise residual signal detected by the noise cancellation system during operation, generated by superimposing the noise-cancelled signal and the in-vehicle noise signal. The error signal is used to evaluate the current noise cancellation effect and guide the system to adjust the noise cancellation strategy to continuously optimize the filter weight coefficients. Controlling the filter weight coefficients is a key parameter in the active noise cancellation system. By adjusting these coefficients, the generation of the noise-cancelled signal can be optimized to cancel out the current in-vehicle noise as much as possible, achieving the best noise cancellation effect.

[0068] Specifically, the process begins by acquiring current reference and error signals through sensors, and then preprocessing these signals, such as filtering, denoising, and normalization. Next, a pre-defined feature extraction algorithm is used to analyze the current signal, extracting parameters that reflect the signal's essential characteristics (such as spectral features and energy distribution). These extracted feature data are then fed into a pre-trained mapping model. This mapping model, trained on historical data, can identify the correlation between input signal features and the optimal control filter weights.

[0069] Furthermore, after receiving the current signal characteristics from the input, the mapping model quickly analyzes and matches the corresponding control filter weight coefficients. The weight coefficients output by the model are the optimal noise reduction coefficients for the current environment. These noise reduction coefficients are used to control the adjustment of the filter, ensuring that the generated noise-reduced signal can cancel out the current noise to the greatest extent possible. Since this process is based on rapid matching of mapping relationships, no complex real-time calculations are required, and the system can quickly adapt to changing noise environments, maintaining optimal noise reduction performance.

[0070] By extracting the control filter weight coefficients corresponding to the current signal based on the mapping relationship, the system can quickly locate the optimal noise reduction coefficients. This model-predictive-based rapid matching significantly shortens the filter adjustment time, enabling the system to react quickly to changes in the noise environment and greatly improving noise reduction efficiency. The use of the mapping relationship ensures a close match between the filter weight coefficients and the current noise characteristics, allowing the system to remain in an optimal state even with frequent changes in the noise environment. This adaptability enables the system to cope with varying noise sources, such as different road conditions, vehicle speed changes, or external interference, ensuring the consistency and stability of the noise reduction effect. By quickly extracting the optimal noise reduction coefficients corresponding to the current environment, the system can generate a noise reduction signal that is inversely phase to the current noise, effectively reducing the error signal inside the vehicle. This precise matching and dynamic adjustment capability makes the in-vehicle noise reduction effect more significant, allowing passengers to enjoy a quieter and more comfortable in-vehicle environment and improving the overall driving experience. This step, through the efficient application of the mapping relationship, achieves rapid and accurate matching of the control filter weight coefficients, greatly improving the response speed and adaptability of the in-vehicle noise reduction system.

[0071] S4: Convolve the noise reduction coefficients with the current reference signal to generate the target noise reduction signal.

[0072] It should be noted that convolution is a mathematical operation that combines two signals to generate a new signal. Its main role in signal processing is filtering; that is, by combining a reference signal with a filter (noise reduction coefficient) through convolution, an output signal can be generated that performs specific processing on the original signal.

[0073] Specifically, first, prepare the extracted noise reduction coefficients and the current reference signal. Use the noise reduction coefficients as the weight input to the filter and the reference signal as the input signal. Before the convolution operation begins, both the reference signal and the filter coefficients must be normalized to ensure signal amplitude compatibility and avoid distortion during the operation.

[0074] Furthermore, the convolution operation generates the output signal by sliding the denoising coefficients as a sliding window across the entire reference signal, calculating a weighted sum at each position. Specifically, at each step of the convolution, the denoising coefficients are multiplied and summed with the corresponding portion of the current reference signal to generate the output value at that position. Through this point-by-point calculation, a complete target denoised signal is ultimately formed. This signal has the opposite phase and adjusted amplitude to the original noise signal, thus effectively canceling noise during physical superposition.

[0075] By combining the noise reduction coefficients with the current reference signal through convolution operations, the generated target noise-reduced signal can accurately cancel out the current noise. This is because the convolution process adjusts the phase and amplitude of the signal, allowing the generated signal to form an inverse waveform with the original noise, thus physically canceling each other out and achieving effective noise reduction. Since the convolution calculation is based on real-time acquired reference signals and dynamically adjusted noise reduction coefficients, the generated target noise-reduced signal can quickly adapt to changes in noise in the in-vehicle environment. Whether it's engine noise, road vibration, or external interference noise, the system can generate a suitable noise reduction signal in real time, ensuring the consistency and stability of the noise reduction effect. After the generated target noise-reduced signal is superimposed inversely with the in-vehicle noise, it effectively reduces the noise level inside the vehicle, creating a quieter and more comfortable environment for passengers. This not only improves the riding experience but also reduces fatigue from prolonged exposure to noise, enhancing the overall comfort of the vehicle and user satisfaction. This step, combining the noise reduction coefficients with the reference signal through convolution operations, achieves precise cancellation of in-vehicle noise and is a key operation in the entire noise reduction system for noise suppression.

[0076] This embodiment generates multiple control filter weight coefficients based on an adaptive filtering algorithm. It then determines the mapping relationship between the preset training reference signal, the preset training error signal, and the control filter weight coefficients based on a preset feature extraction algorithm. Based on this mapping relationship, it extracts the control filter weight coefficients corresponding to the current reference signal and the current error signal, using them as noise reduction coefficients. Finally, it convolves these noise reduction coefficients with the current reference signal to generate the target noise-reduced signal. This embodiment utilizes an adaptive filtering algorithm and feature extraction mapping technology to achieve rapid matching and dynamic adjustment of the filter weight coefficients, effectively improving noise reduction speed and applicability. This allows the in-vehicle active noise cancellation system to maintain efficient and stable noise reduction performance even in complex and variable noise environments.

[0077] Based on the first embodiment described above, a second embodiment of the in-vehicle noise reduction method of this application is proposed. Please refer to... Figure 2 , Figure 2 This is a schematic diagram of a sub-process in the second embodiment of the in-vehicle noise reduction method of this application.

[0078] like Figure 2 As shown, in this embodiment, step S1 includes:

[0079] S11: Use noise information as input to the preset filter;

[0080] S12: Initialize the weight coefficients and length of the preset filter;

[0081] S13: Based on the adaptive filtering algorithm, adjust the weight coefficients of the preset filter to obtain the weight coefficients of the plurality of control filters.

[0082] It should be noted that the preset filter is a dynamic filter used for signal processing. Its weight coefficients can be continuously adjusted according to the input signal to optimize the filtering effect. The preset filter is usually initialized with zero or a small value, and dynamically adjusted through an adaptive algorithm as the system runs to achieve optimal processing of noisy signals.

[0083] Specifically, noise information is collected from sensors inside the vehicle and used as the input signal for a preset filter. This noise information can include noise characteristics from different sources such as the engine, road surface, and wind noise. Next, the weight coefficients and length of the preset filter are initialized. The initial values ​​of the weight coefficients are generally set to zero or random small values ​​close to zero, while the filter length is set according to the system design requirements and the complexity of the input noise. This initialization ensures that the filter does not interfere excessively with noise in its initial state, awaiting further adjustments by the adaptive algorithm.

[0084] Furthermore, after filter initialization, the adaptive filtering algorithm begins to dynamically adjust the preset filter. Through this adaptive algorithm, the system continuously adjusts the weight coefficients based on real-time acquired noise information and the filter's output error signal. The goal of the adjustment is to minimize the error signal, i.e., the difference between the filter output and the desired output. The algorithm iteratively optimizes and gradually adjusts the weight coefficients. After multiple iterations, the preset filter can generate multiple control filter weight coefficients. These weight coefficients represent the optimal set of control parameters for different noise environments and can be used for rapid matching and application in subsequent noise reduction processes.

[0085] By dynamically adjusting the weight coefficients of the preset filter using an adaptive algorithm, the system can quickly adapt to different noise environments and generate multiple control filter weight coefficients. This adjustment mechanism enables the filter to respond promptly to noise changes, improving the efficiency of noise reduction signal generation and reducing system response delay. The automatic adjustment of the preset filter weight coefficients under different noise conditions allows the system to flexibly handle various noise sources, such as engine noise, road vibration noise, and external interference noise. Different sets of control weight coefficients can cover various noise modes, enhancing the system's adaptability and maintaining optimal noise reduction performance even in complex and changing environments. The adaptive algorithm aims to minimize the error signal, meaning the generated control filter weight coefficients are an optimized set of parameters. This optimization allows the filter to more accurately generate a noise reduction signal that is the opposite of the noise, effectively reducing in-vehicle noise levels and providing a quieter and more comfortable environment for passengers. This step, through the initialization and adaptive adjustment of the preset filter, generates multiple control filter weight coefficients, providing a flexible parameter basis for subsequent noise reduction processes and improving the overall noise reduction performance of the system.

[0086] Based on the first embodiment described above, in this embodiment, step S2 includes:

[0087] S21: Preprocess the preset training reference signal and the preset training error signal to obtain a preprocessed signal, wherein the preprocessing includes one or more of DC component removal processing, noise removal processing and normalization processing;

[0088] S22: Based on the preset feature extraction algorithm, feature extraction is performed on the preprocessed signal to obtain noise features;

[0089] S23: Combine the noise features with the corresponding control filter weight coefficients to obtain the mapping relationship.

[0090] It should be noted that the preset training reference signal is a reference signal sample collected from historical data or experimental environments, used to train and optimize the control filter of the noise reduction system. This signal typically reflects the main characteristics of the noise source, providing a basis for establishing the mapping relationship between features and filter weight coefficients. The preset training error signal is a residual noise signal that accompanies the reference signal, used to measure the noise reduction effect and guide the optimization of filter weight coefficients. It represents the difference between the noise-reduced signal and the actual noise, serving as feedback data during the control filter adjustment process. Feature extraction is the process of transforming the preprocessed signal into feature data that is easier to analyze and model. Through feature extraction, key information in the signal, such as spectral characteristics and energy distribution, can be extracted, providing a basis for establishing the mapping relationship between noise features and control filter weight coefficients. The mapping relationship refers to the combination of signal features and corresponding control filter weight coefficients through feature extraction algorithms, thereby establishing a correlation model between input signal features and output weight coefficients. This relationship enables the system to quickly select and adjust weight coefficients to optimize the noise reduction effect during actual operation.

[0091] Specifically, preprocessing is performed on the preset training reference signal and error signal to improve signal quality and consistency. Preprocessing includes DC component removal, noise removal, and normalization. DC component removal eliminates signal offset by subtracting the signal's average value, ensuring signal symmetry around zero. Noise removal removes unwanted background noise using methods such as filtering, improving clarity. Normalization adjusts the signal amplitude to a uniform range for subsequent feature extraction and model training. These processing steps eliminate interference factors in the signal, ensuring the system's stability and robustness under different signal conditions.

[0092] Furthermore, after preprocessing, the system uses a preset feature extraction algorithm (such as Fourier transform, short-time Fourier transform, or Mel-frequency analysis) to extract features from the preprocessed signal. The goal of feature extraction is to transform the signal into a data form that accurately describes its essential characteristics, such as spectral energy, frequency distribution, and time-frequency structure. These noise features are combined with the corresponding control filter weights to form training samples. Through these samples, the system can train a mapping model that can identify the relationship between features and weights, thereby quickly selecting the optimal control filter weights based on the current noise characteristics during actual noise reduction.

[0093] By preprocessing the preset training signal, the system effectively removes DC components and noise interference, ensuring the input training signal has good quality and stability. Signal normalization ensures the feature extraction algorithm operates within a uniform amplitude range, reducing extraction bias caused by differences in signal amplitude. The preprocessed signal more accurately reflects the essential characteristics of noise during feature extraction. These accurate feature data, combined with corresponding weight coefficients, enable the mapping model to better learn the relationship between features and weight coefficients, thus providing more precise noise reduction parameter selection in practical applications. By establishing a mapping relationship between preset training signal features and control filter weight coefficients, the system can quickly identify the current noise environment and select weight coefficients, significantly improving the system's adaptability and noise reduction effect. Even in complex and variable noise scenarios, the system can dynamically adjust the weight coefficients, maintaining the stability and consistency of noise reduction performance. This step, through signal preprocessing, feature extraction, and the establishment of the mapping relationship, lays a solid foundation for the dynamic adjustment of the control filter, giving the noise reduction system higher accuracy and adaptability.

[0094] This embodiment generates multiple control filter weight coefficients based on an adaptive filtering algorithm. It then determines the mapping relationship between the preset training reference signal, the preset training error signal, and the control filter weight coefficients based on a preset feature extraction algorithm. Based on this mapping relationship, it extracts the control filter weight coefficients corresponding to the current reference signal and the current error signal, using them as noise reduction coefficients. Finally, it convolves these noise reduction coefficients with the current reference signal to generate the target noise-reduced signal. This embodiment utilizes an adaptive filtering algorithm and feature extraction mapping technology to achieve rapid matching and dynamic adjustment of the filter weight coefficients, effectively improving noise reduction speed and applicability. This allows the in-vehicle active noise cancellation system to maintain efficient and stable noise reduction performance even in complex and variable noise environments.

[0095] Based on the second embodiment described above, a third embodiment of the in-vehicle noise reduction method of this application is proposed. Please refer to... Figure 3 , Figure 3 This is a schematic diagram of a sub-process in the third embodiment of the in-vehicle noise reduction method of this application.

[0096] In this embodiment, after step S23, the following is also included:

[0097] S23a: Construct a mapping model based on the noise characteristics, the weight coefficients of the multiple control filters, and the mapping relationship;

[0098] S23b: Evaluate the training performance of the mapping model using a loss function;

[0099] S23c: Validate the mapping model that has passed the evaluation, and adjust the hyperparameters of the mapping model based on the validation results.

[0100] It's important to note that noise features are key information extracted from the pre-set training signal, reflecting the main characteristics of the noise signal, such as spectral features, energy distribution, and time-frequency domain features. These features are used to establish a correlation with the control filter weights to optimize the noise reduction effect. The mapping model is a data-driven model (such as neural networks or regression models) used to predict the optimal filter weights under the current noise environment. This model outputs corresponding filter weights based on the noise features, achieving intelligent control and adaptive adjustment of the system. The loss function is a standard used to evaluate the training effect of the mapping model. It measures the difference between the model's predicted output and the true value. Common loss functions include mean squared error (MSE) and cross-entropy loss. By minimizing the loss function value, the training results of the model are optimized. Hyperparameters are adjustable parameters during model training, such as the learning rate, the number of neural network layers, and the number of neurons. Different hyperparameter settings affect the model's training effect and prediction performance. Optimizing hyperparameters helps improve the model's generalization ability and prediction accuracy.

[0101] Specifically, through the aforementioned feature extraction and mapping relationship establishment, the system combines noise features with multiple control filter weight coefficients as training data to construct a mapping model. The mapping model can be constructed using machine learning or deep learning techniques such as neural networks and support vector regression (SVR). During training, the model continuously adjusts its internal parameters to learn the relationship between noise features and filter weight coefficients. After training, the model can output the corresponding optimal control filter weight coefficients based on the input noise features, providing support for the subsequent noise reduction process.

[0102] Furthermore, after the mapping model is trained, the system evaluates the model's predictive performance using a loss function. The loss function calculates the error between the model's predicted output and the true weight coefficients; for example, the mean squared error (MSE) measures the difference between the predicted and actual values. After evaluation, the system validates the model's performance and adjusts hyperparameters based on the validation results, such as adjusting the learning rate, the number of neural network layers, and the batch size. Optimizing hyperparameters can improve the model's training performance, reduce the risk of overfitting, and enhance the model's generalization ability under different noise environments.

[0103] By constructing, evaluating, and optimizing a mapping model, the system can accurately predict the optimal control filter weights corresponding to the current noise characteristics. This high-precision mapping model enables the noise reduction system to respond quickly in complex and variable noise environments, significantly improving the system's adaptability and noise reduction effect. By evaluating the mapping model using a loss function, the system can identify the sources of error in the model and optimize it by adjusting hyperparameters. The optimized mapping model can more accurately match noise characteristics with filter weights, reduce prediction errors, and make the generated noise-reduced signal more accurate, significantly improving the in-vehicle noise suppression effect. Through validation and hyperparameter tuning, the system ensures that the mapping model not only performs well on training data but also effectively handles different noise scenarios in practical applications. This optimization process improves the model's generalization ability, enabling the system to maintain efficient noise reduction in various complex noise environments, ensuring the consistency and stability of the user experience. This step, through constructing, evaluating, and optimizing the mapping model, achieves accurate matching between noise characteristics and control filter weights, providing the noise reduction system with core intelligent adjustment capabilities.

[0104] Based on the second embodiment described above, in this embodiment, step S3 includes:

[0105] S31: Perform the preprocessing on the current reference signal and the current error signal to obtain the current processed signal;

[0106] S32: Based on the preset feature extraction algorithm, perform feature extraction on the currently processed signal to obtain the current feature matrix;

[0107] S33: Input the current feature matrix into the mapping model to obtain the corresponding control filter weight coefficients, and use them as the noise reduction coefficients.

[0108] It should be noted that the current reference signal is the noise signal collected in real time by the active noise cancellation system during actual operation, reflecting the characteristics of noise sources in the current in-vehicle environment, including engine noise, road noise, and wind noise. These signals are the basic inputs for noise cancellation processing, used to generate a noise-reduced signal that is the opposite of the noise. The current error signal is the actual noise signal remaining after the system generates the noise-reduced signal. It represents feedback information on the noise reduction effect and is used to adjust and optimize the weight coefficients of the control filter to further reduce in-vehicle noise. The feature matrix is ​​the result after processing by the feature extraction algorithm, containing multi-dimensional feature data of the current noise signal. This matrix structurally displays the time-frequency characteristics of the signal, providing rich information input for the mapping model. The mapping model is a predictive model trained based on the correlation between noise features and control filter weight coefficients, which can output the corresponding optimal control filter weight coefficients according to the input noise features to optimize the noise reduction effect.

[0109] Specifically, the system first preprocesses the current reference signal and the current error signal, removing the DC component, eliminating noise interference, and normalizing them to obtain a clearer and more stable processed signal. This processing ensures the consistency and high quality of the input signal. Next, the system uses preset feature extraction algorithms (such as Fourier transform and short-time Fourier transform) to extract features from the processed signal, transforming it into a multi-dimensional feature matrix. This feature matrix captures key information such as the frequency, amplitude, and energy of the current noise signal, accurately reflecting the essential characteristics of the noise.

[0110] Furthermore, after feature extraction, the current feature matrix is ​​input into the previously constructed mapping model. This mapping model is trained based on the mapping relationship between noise features and control filter weights, enabling it to quickly analyze the input feature matrix and predict the optimal control filter weights. These weights output by the model are the optimal denoising coefficients for the current environment, used to control the filter to generate a denoised signal opposite to the noise. This real-time prediction and application mechanism ensures that the denoising system maintains efficient and stable denoising performance in varying noise environments.

[0111] Through preprocessing and feature extraction, the system can quickly acquire high-quality feature data of the current noise environment and directly input these features into the mapping model for weight coefficient prediction. This process reduces computational latency, enables rapid response to noise changes, and improves the real-time performance and adaptability of the noise reduction system. The mapping model can quickly match the optimal control filter weight coefficients based on the input feature matrix, generating accurate noise reduction coefficients. These optimized weight coefficients ensure that the noise reduction signal generated by the filter has the best anti-matching effect with the current noise, effectively reducing residual noise inside the vehicle and significantly improving the noise reduction effect. Through precise analysis of the feature matrix and real-time adjustment of the weight coefficients, the system can continuously optimize the output signal of the filter and reduce the amplitude of the error signal. This process ensures that the noise reduction system can operate stably under various noise conditions, providing passengers with a quieter in-vehicle environment. This step, through signal preprocessing, feature extraction, and intelligent prediction by the mapping model, achieves rapid adaptation to the noise environment and accurate noise reduction, providing the core real-time adjustment capability for the in-vehicle noise reduction system.

[0112] Based on the second embodiment described above, in this embodiment, step S4 includes:

[0113] S41: Align the initial position of the noise reduction coefficient with the starting point of the current reference signal;

[0114] S42: Determine the weighted sum at each position by moving the noise reduction coefficients point by point along the current reference signal;

[0115] S43: Generate a noise-reduced signal based on the weighted sum;

[0116] S44: Verify the denoised signal based on the current error signal, and use the verified denoised signal as the target denoised signal.

[0117] It should be noted that the weighted sum refers to the sum of the products of the noise reduction coefficients and each point of the current reference signal. In convolution operations, the weighted sum represents the superposition effect of the signal and filter coefficients at each position, and is a crucial step in generating the noise-reduced signal. The target noise-reduced signal is a verified and effective noise-reduced signal that can significantly reduce in-vehicle noise and provide a quiet environment for passengers. This signal represents the final noise reduction effect output by the system.

[0118] Specifically, the initial positions of the denoising coefficients are aligned with the starting point of the current reference signal to ensure that the convolution operation begins from the signal's origin. Then, by moving the denoising coefficients point-by-point along the current reference signal, the system calculates the product of the denoising coefficient and the signal value at each position and calculates a weighted sum. Specifically, at each convolution position, the denoising coefficient is multiplied one by one by a subsequence of the current reference signal, and the sum is the weighted sum, which is the output value at that position. Through point-by-point movement and cumulative calculation, a complete denoised signal is finally generated, which is the inverse cancellation result of the current noise signal.

[0119] Furthermore, after generating the denoised signal, the system verifies it based on the current error signal. By analyzing the magnitude and trend of the error signal, the system determines whether the generated denoised signal is effective, i.e., whether it can significantly reduce noise. If the verification passes, it indicates that the denoised signal has a good cancellation effect with the current noise signal, and this signal is confirmed and output as the final target denoised signal. If the verification fails, the system may need to adjust the denoising coefficient or recalculate to ensure that the final output target signal achieves the expected denoising effect.

[0120] By calculating the weighted sum of the noise reduction coefficients and the reference signal point by point, the system can generate a noise-reduced signal that highly matches the current noise signal. This precise calculation method ensures that the phase and amplitude of the noise-reduced signal are opposite to those of the noise signal, effectively improving the noise cancellation effect and reducing residual noise inside the vehicle. By verifying the generated noise-reduced signal with error signals, the system can dynamically adjust the noise reduction strategy to ensure that the output signal remains effective under different noise environments. This verification mechanism enhances the stability of the system, making the noise reduction effect more reliable and consistent in practical applications. The generation and verification process of the target noise-reduced signal ensures the system's precise control and dynamic adjustment of in-vehicle noise, providing passengers with a quiet and comfortable in-vehicle environment. This not only improves the driving and riding experience but also reduces fatigue and discomfort caused by long-term exposure to noise. This step, through point-by-point convolution calculation and dynamic verification, ensures that the generated noise-reduced signal can accurately and effectively cancel the current noise, and is a key step in achieving efficient noise reduction.

[0121] This embodiment generates multiple control filter weight coefficients based on an adaptive filtering algorithm. It then determines the mapping relationship between the preset training reference signal, the preset training error signal, and the control filter weight coefficients based on a preset feature extraction algorithm. Based on this mapping relationship, it extracts the control filter weight coefficients corresponding to the current reference signal and the current error signal, using them as noise reduction coefficients. Finally, it convolves these noise reduction coefficients with the current reference signal to generate the target noise-reduced signal. This embodiment utilizes an adaptive filtering algorithm and feature extraction mapping technology to achieve rapid matching and dynamic adjustment of the filter weight coefficients, effectively improving noise reduction speed and applicability. This allows the in-vehicle active noise cancellation system to maintain efficient and stable noise reduction performance even in complex and variable noise environments.

[0122] Please see Figures 4-6 , Figure 4 A flowchart for establishing the set of noise reduction weights and training the CNN model; Figure 5 This is a schematic diagram showing the sub-band division of the reference signal and error signal; Figure 6 This is a block diagram of a 2D CNN model.

[0123] In one embodiment, the in-vehicle noise reduction method includes:

[0124] The reference signal x(n) passes through the secondary path The filtered reference signal x′(n) is obtained after the filtering module.

[0125]

[0126] The filtered reference signal x′(n) is decomposed into m sub-bandpass signals x′ after passing through bandpass filter bank 1. i (n), i = 1, 2…m. Bandpass filter: i = 1, 2, ..., m (m is an even number).

[0127] The error signal e(n) is decomposed into m sub-band signals e after passing through bandpass filter bank 2. i (n), i = 1, 2…m, where the bandwidth and frequency range of each sub-band are related to the sub-band vibration reference signal x. i (n) are the same.

[0128] The order of the noise reduction control filter W(n) is set to N, and each sub-band noise reduction control filter W i order N i = N / m.

[0129] Sub-band reference signal x′ i (n) and error signal e i (n) Input the weight coefficients W of the adaptive noise reduction control filter for each sub-band i (n) The optimal system is found by using the Normalized Least Mean Square (NLMS) algorithm to solve for the optimal weight coefficients of each sub-band noise reduction control filter:

[0130]

[0131] Among them: W i (n+1) - the current control filter weight coefficients, W i (n) - Weight coefficients of the control filter at the previous time step, x′ i (n) - Reference signal, -Reference signal transpose, e i (n) - error signal, α - convergence factor, β - constant.

[0132] Weight coefficients W of the sub-band adaptive noise reduction control filter i (n) Perform a multi-point Fourier transform (FFT) on the subband adaptive noise reduction control filter weight coefficients W. i (n) Transformation from the time domain to the frequency domain W i (z).

[0133] W i (z)=FFT(W i (n))

[0134] In the frequency domain, the weighting coefficients W of the subband adaptive noise reduction control filter are... i Stack the (z) values ​​to obtain the weight coefficients W(z) of the full-band adaptive noise reduction control filter in the frequency domain.

[0135]

[0136] The weight coefficients W(z) of the full-band adaptive noise reduction control filter in the frequency domain are obtained by performing an inverse Fourier transform (IFFT) to obtain the weight coefficients W(n) of the full-band adaptive noise reduction control filter in the time domain.

[0137] W(n) = IFFT(W(z))

[0138] The reference vibration signal x(n) is convolved with the weight coefficients W(n) of the time-domain full-band adaptive noise reduction control filter to generate the output signal u(n) of the noise reduction control filter.

[0139] u(n) = x(n) * W(n)

[0140] u(n) emits a noise-reducing wave y(n) through the secondary path S(n) (which includes digital-to-analog conversion (DA), power amplification (AMP), etc.), which is superimposed on the original sound d(n) inside the vehicle in the opposite phase, thereby achieving the purpose of eliminating / reducing the noise inside the vehicle.

[0141] e(n) = d(n) - y(n)

[0142] After the system achieves stable noise reduction, 2D CNN model training is initiated; the weight coefficients W of the noise reduction control filter for each subband are adjusted. i Tagging (r) i ,s i →W i (n); for each sub-band reference signal x′ i (n) and error signal e i (n) Perform frame segmentation; solve for the Mel feature spectrum of the signal in each frame, and use it as the input to the 2D CNN network. Use the corresponding sub-band noise reduction control filter weight coefficient labels as the output, and establish the mapping relationship (x′) between the input layer and the output layer. i (n),e i (n))→(r i ,s i ).

[0143] Repeat the above steps on different road surfaces to establish a set of weight coefficients for the sub-band noise reduction control filter under different road surfaces, and the mapping relationship between the corresponding reference signal, error signal and the weight coefficient labels of the output sub-band noise reduction control filter.

[0144] After the model training is completed, it is applied to the noise reduction process. The reference signal and the error signal are divided into sub-bands, and the signals of each sub-band are framed. The Mel feature spectrum of each frame signal is solved. The solved Mel feature spectrum is output to the input layer of the 2D CNN network. The 2D CNN network identifies the Mel feature spectrum of the reference signal and the error signal. According to the mapping relationship established earlier, it outputs the label of the corresponding sub-band noise reduction control filter weight coefficient. The corresponding sub-band noise reduction control filter weight coefficient is found and extracted by the label.

[0145] (x′ i (n),e i (n))→(r i ,s i →W i (n)

[0146] The weighting coefficients W of the extracted subband noise reduction control filter i (n) Perform a multi-point Fourier transform (FFT) on the subband noise reduction control filter weight coefficients W. i (n) Transformation from the time domain to the frequency domain W i (z).

[0147] W i (z)=FFT(W i (n))

[0148] In the frequency domain, the weighting coefficients W of the subband noise reduction control filter are... i Stack the (z) values ​​to obtain the weight coefficients W(z) of the full-band noise reduction control filter in the frequency domain.

[0149]

[0150] The inverse Fourier transform (IFFT) is performed on the weight coefficients W(z) of the full-band noise reduction control filter in the frequency domain to obtain the weight coefficients W(n) of the full-band noise reduction control filter in the time domain.

[0151] W(n) = IFFT(W(z))

[0152] The reference vibration signal x(n) is convolved with the weight coefficients W(n) of the time-domain full-band noise reduction control filter to obtain the output signal u(n) of the noise reduction control filter.

[0153] u(n) = x(n) * W(n)

[0154] u(n) emits a noise-reducing wave y(n) through the secondary path S(n) (which includes digital-to-analog conversion (DA), power amplification (AMP), etc.), which is superimposed on the original sound d(n) inside the vehicle in the opposite phase, thereby achieving the purpose of eliminating / reducing the noise inside the vehicle.

[0155] e(n) = d(n) - y(n)

[0156] Sub-band reference signal x′ i (n) and error signal e i (n) Input the weight coefficients W of the adaptive noise reduction control filter for each sub-band i (n) The optimal system is found by using the Normalized Least Mean Square (NLMS) algorithm to solve for the optimal noise reduction control filter weights for each sub-band:

[0157]

[0158] Among them: W i (n+1) - Control filter coefficients at the current time, W i (n) - Control filter coefficients from the previous time step, x′ i (n) - Reference signal, -Reference signal transpose, e i (n) - error signal, α - convergence factor, β - constant.

[0159] The weight coefficients W of the subband noise reduction control filter are obtained by solving. i (n) Replace the weight coefficients in the corresponding control filter weight coefficient set library and correct the weight coefficients.

[0160] This application also provides an in-vehicle noise reduction device; please refer to... Figure 7 , Figure 7 This is a schematic diagram of the module structure of the in-vehicle noise reduction device according to an embodiment of this application. The in-vehicle noise reduction device includes:

[0161] The coefficient determination module 701 is used to generate multiple control filter weight coefficients based on an adaptive filtering algorithm;

[0162] The mapping determination module 702 is used to determine the mapping relationship between the preset training reference signal and the preset training error signal and the weight coefficients of the control filter based on the preset feature extraction algorithm.

[0163] The coefficient extraction module 703 is used to extract the control filter weight coefficients corresponding to the current reference signal and the current error signal based on the mapping relationship, and use them as noise reduction coefficients.

[0164] The target module 704 is used to convolve the noise reduction coefficients with the current reference signal to generate a target noise reduction signal.

[0165] The in-vehicle noise reduction device provided in this application adopts the in-vehicle noise reduction method in the above embodiments, which can solve the technical problem of how to improve the speed and applicability of in-vehicle noise reduction. Compared with the prior art, the beneficial effects of the in-vehicle noise reduction device provided in this application are the same as the beneficial effects of the in-vehicle noise reduction method provided in the above embodiments, and other technical features in the in-vehicle noise reduction device are the same as the features disclosed in the methods of the above embodiments, and will not be repeated here.

[0166] This application provides an in-vehicle noise reduction device, which includes: 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 perform the in-vehicle noise reduction method in the above embodiments.

[0167] The following is for reference. Figure 8 The diagram illustrates a structural schematic suitable for implementing an in-vehicle noise reduction device according to embodiments of this application. The in-vehicle noise reduction device in embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 8 The in-vehicle noise reduction device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of this application.

[0168] like Figure 8 As shown, the in-vehicle noise reduction device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the in-vehicle noise reduction device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the in-vehicle noise reduction equipment to communicate wirelessly or wiredly with other devices to exchange data. Although in-vehicle noise reduction equipment with various systems is shown in the figures, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems may be implemented alternatively.

[0169] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0170] The in-vehicle noise reduction device provided in this application, employing the in-vehicle noise reduction method described in the above embodiments, can solve the technical problem of how to improve the speed and applicability of in-vehicle noise reduction. Compared with the prior art, the beneficial effects of the in-vehicle noise reduction device provided in this application are the same as those of the in-vehicle noise reduction method provided in the above embodiments, and other technical features of the in-vehicle noise reduction device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0171] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0172] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0173] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the in-vehicle noise reduction method in the above embodiments.

[0174] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof. The aforementioned computer-readable storage medium may be included in the in-vehicle noise reduction device; or it may exist independently and not be installed in the in-vehicle noise reduction device.

[0175] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by the in-vehicle noise reduction device, the in-vehicle noise reduction device: generates multiple control filter weight coefficients based on an adaptive filtering algorithm; determines the mapping relationship between a preset training reference signal and a preset training error signal and the control filter weight coefficients based on a preset feature extraction algorithm; extracts the control filter weight coefficients corresponding to the current reference signal and the current error signal based on the mapping relationship, and uses them as noise reduction coefficients; and convolves the noise reduction coefficients with the current reference signal to generate a target noise reduction signal. Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on a user's computer, partially on a user's computer, as a standalone software package, partially on a user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer through any type of network—including a local area network (LAN) or a wide area network (WAN)—or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0176] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0177] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0178] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., computer programs) for executing the above-described in-vehicle noise reduction method, thereby solving the technical problem of how to improve the speed and applicability of in-vehicle noise reduction. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the in-vehicle noise reduction method provided in the above embodiments, and will not be repeated here.

[0179] This application provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the in-vehicle noise reduction method described above.

[0180] The computer program product provided in this application can solve the technical problem of how to improve the speed and applicability of in-vehicle noise reduction. Compared with the prior art, the beneficial effects of the computer program product provided in the embodiments of this application are the same as the beneficial effects of the in-vehicle noise reduction method provided in the above embodiments, and will not be repeated here.

[0181] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent scope of this application.

Claims

1. A method for reducing noise inside a vehicle, characterized in that, The method includes: Based on an adaptive filtering algorithm, multiple control filter weight coefficients are generated. The preset training reference signal and the preset training error signal are preprocessed to obtain the preprocessed signal; Based on a preset feature extraction algorithm, feature extraction is performed on the preprocessed signal to obtain noise features; The noise features are combined with the corresponding control filter weight coefficients to obtain the mapping relationship; Based on the noise features, multiple control filter weights, and the mapping relationship, a mapping model is constructed, wherein the mapping model is a 2D CNN model. By dividing the reference signal and error signal of each sub-band of the mapping model into frames, the Mel feature spectrum of each frame signal is determined and used as the input of the 2D CNN network. The weights of the noise reduction control filters of each sub-band of the mapping model are labeled, and the corresponding labels of the sub-band noise reduction control filter weights are used as the output of the 2D CNN network. The training performance of the mapping model is evaluated using a loss function; The mapping model that passes the evaluation is validated, and the hyperparameters of the mapping model are adjusted based on the validation results; Based on the mapping relationship, the control filter weight coefficients corresponding to the current reference signal and the current error signal are extracted and used as noise reduction coefficients; The noise reduction coefficients are convolved with the current reference signal to generate the target noise-reduced signal.

2. The method as described in claim 1, characterized in that, The step of generating multiple control filter weight coefficients based on the adaptive filtering algorithm includes: Use noise information as input to a preset filter; Initialize the weight coefficients and length of the preset filter; Based on the adaptive filtering algorithm, the weight coefficients of the preset filter are adjusted to obtain the weight coefficients of the plurality of control filters.

3. The method as described in claim 1, characterized in that, The step of extracting the control filter weight coefficients corresponding to the current reference signal and the current error signal based on the mapping relationship, and using them as noise reduction coefficients, includes: The preprocessing is performed on the current reference signal and the current error signal to obtain the current processed signal; Based on the preset feature extraction algorithm, features are extracted from the currently processed signal to obtain the current feature matrix; The current feature matrix is ​​input into the mapping model to obtain the corresponding control filter weight coefficients, which are then used as the noise reduction coefficients.

4. The method as described in claim 1, characterized in that, The step of convolving the noise reduction coefficients with the current reference signal to generate the target noise-reduced signal includes: Align the initial position of the noise reduction coefficient with the starting point of the current reference signal; The weighted sum at each position is determined by moving the noise reduction coefficients point by point along the current reference signal; Based on the weighted sum, a noise-reduced signal is generated; The denoised signal is verified based on the current error signal, and the verified denoised signal is taken as the target denoised signal.

5. A vehicle interior noise reduction device, characterized in that, The device includes: The coefficient determination module is used to generate multiple control filter weight coefficients based on an adaptive filtering algorithm. A mapping determination module is used to preprocess a preset training reference signal and a preset training error signal to obtain a preprocessed signal; extract features from the preprocessed signal based on a preset feature extraction algorithm to obtain noise features; combine the noise features with the corresponding control filter weight coefficients to obtain a mapping relationship; construct a mapping model based on the noise features, multiple control filter weight coefficients, and the mapping relationship, wherein the mapping model is a 2D CNN model; divide the sub-band reference signal and error signal of the mapping model into frames, determine the Mel feature spectrum of each frame signal and use it as the input of the 2D CNN network; label the weight coefficients of each sub-band noise reduction control filter of the mapping model, and use the corresponding sub-band noise reduction control filter weight coefficient labels as the output of the 2D CNN network; evaluate the training effect of the mapping model through a loss function; validate the mapping model that passes the evaluation, and adjust the hyperparameters of the mapping model based on the validation results; The coefficient extraction module is used to extract the control filter weight coefficients corresponding to the current reference signal and the current error signal based on the mapping relationship, and use them as noise reduction coefficients. The target module is used to convolve the noise reduction coefficients with the current reference signal to generate a target noise reduction signal.

6. A computer device, characterized in that, The device includes: a memory, a processor, and an in-vehicle noise reduction program stored in the memory and executable on the processor, the in-vehicle noise reduction program being configured to implement the steps of the in-vehicle noise reduction method as described in any one of claims 1 to 4.

7. A storage medium, characterized in that, The storage medium stores an in-vehicle noise reduction program, which, when executed by a processor, implements the steps of the in-vehicle noise reduction method as described in any one of claims 1 to 4.

8. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the in-vehicle noise reduction method as described in any one of claims 1 to 4.

Images