Vehicle noise reduction method, device, storage medium and equipment

Abstract

The present application discloses a vehicle noise reduction method, device, storage medium and equipment. The method comprises: firstly, inputting a target road noise signal generated by a target vehicle into a pre-constructed fixed filter selection model, and predicting a target fixed filter for noise reduction of the target vehicle; wherein the target fixed filter is determined from N fixed filters pre-installed inside the target vehicle; the parameters of the N fixed filters are determined by performing optimal filter calculation on the road noise signals under N typical working conditions collected in advance; the fixed filter selection model is obtained by jointly training the road noise signals under N working conditions and the N fixed filters; then, the target road noise signal is filtered by using the target fixed filter, and the anti-phase noise control signal obtained after filtering is played through the in-vehicle speaker to interfere with and cancel the target road noise signal, thereby improving the noise reduction effect on the target vehicle and reducing the filter selection cost.

Description

Technical Field

[0001] The present application relates to the field of vehicle technology, and in particular to a vehicle noise reduction method, device, storage medium and equipment. Background Art

[0002] With the improvement of people's living standards and the rapid development of social economy, the use rate of cars has gradually increased, bringing great convenience to all aspects of people's lives. People's requirements for riding experience are also getting higher and higher. However, the road noise generated by the vehicle during driving will reduce the riding experience of the vehicle users. For example, it may affect the emotions and physical health of the vehicle users, causing hearing damage, irritability, fatigue, etc., and even affect the perception of the vehicle users to the sound outside the car, affecting driving safety. Therefore, reducing road noise is of great significance to the riding comfort of vehicle users and protecting the health of passengers.

[0003] At present, the denoising method for road noise is usually implemented by adaptive filtering algorithm, but this adaptive filtering algorithm has several disadvantages: first, it is easy to diverge, which may lead to ineffective noise elimination and potential instability; second, it has a slow convergence speed, which limits its ability to respond quickly to dynamic or rapidly changing noise environments, resulting in reduced noise reduction effect; third, it is difficult to achieve denoising for complex scenes, because for complex scenes, the algorithm needs to be calibrated and adjusted to adapt to various noise conditions, which makes the implementation process of the algorithm more complicated; fourth, because the adaptive filtering algorithm requires real-time calculation, it also leads to a large amount of calculation in the denoising process, resulting in high calculation load and hardware cost. It can be seen that the existing denoising method for road noise does not achieve the ideal denoising effect, and the denoising cost is high. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to provide a vehicle noise reduction method, device, storage medium and equipment, which can effectively improve the vehicle noise reduction effect and reduce the noise reduction cost.

[0005] The embodiment of the present application provides a vehicle noise reduction method, comprising:

[0006] Acquire a target road noise signal during the driving process of the target vehicle;

[0007] The target road noise signal is input into a pre-built fixed filter selection model to predict a target fixed filter for reducing noise on the target vehicle; wherein the target fixed filter is determined from N fixed filters pre-installed inside the target vehicle; the parameters of the N fixed filters are determined by performing optimal filter calculation on the road noise signals under N typical working conditions collected in advance; the fixed filter selection model is obtained by jointly training the road noise signals under N working conditions collected in advance and the N fixed filters; and N is a positive integer greater than 0;

[0008] The target road noise signal is filtered by using the target fixed filter to obtain an anti-phase noise control signal; and the anti-phase noise control signal is played through the in-vehicle speaker of the target vehicle, so that the target road noise signal and the anti-phase noise control signal interfere with each other to achieve active noise reduction of the target vehicle.

[0009] In a possible implementation, the parameters of the N fixed filters are determined as follows:

[0010] Using an acceleration sensor pre-installed in the sample vehicle to obtain vibration and acceleration signals generated when the sample vehicle is driving under N typical working conditions; and using a microphone pre-installed in the sample vehicle to obtain a vehicle interior noise signal generated when the sample vehicle is driving under N typical working conditions; and using the vibration and acceleration signals and the vehicle interior noise signal to form a sample road noise signal;

[0011] Cleaning the sample road noise signal to obtain a cleaned sample road noise signal;

[0012] The Wiener filter calculation method is used to calculate the optimal filter for the cleaned sample road noise signal, so that the noise reduction amount of the sample road noise signal is maximized, so as to determine the parameters of N fixed filters.

[0013] In a possible implementation, the fixed filter selection model is constructed as follows:

[0014] Filtering the vibration and acceleration signals by using the N fixed filters, and playing the filtered signals through a speaker in a sample vehicle, so that a microphone pre-installed in the sample vehicle collects the signal played by the speaker as a sample control signal;

[0015] Calculate the value of residual noise by using the sample control signal and the in-vehicle noise signal; and determine the Kth fixed filter corresponding to the minimum residual noise value as the filter label corresponding to the vibration and acceleration signals; K is a positive integer greater than 0 and not greater than N;

[0016] The vibration and acceleration signals and their corresponding filter labels and target loss function are used to train an initial fixed filter selection model to obtain the fixed filter selection model.

[0017] In a possible implementation, the initial fixed filter selection model includes two one-dimensional convolutional layers and a linear layer; the initial fixed filter selection model is trained using the vibration and acceleration signals and their corresponding filter labels and target loss functions to obtain the fixed filter selection model, including:

[0018] Inputting the vibration and acceleration signals into two one-dimensional convolutional layers and a linear layer in an initial fixed filter selection model, predicting an Mth fixed filter for denoising the sample vehicle; M is a positive integer greater than 0 and not greater than N;

[0019] Determine whether the values ​​of M and K are consistent. If so, train to obtain the fixed filter selection model; if not, reselect the vibration and acceleration signals to train the initial fixed filter selection model, and during the training process, use the back propagation algorithm to optimize the value of the target loss function calculated using the output result of the model, so as to use the optimized value to update the initial fixed filter selection model for training, until the optimized value meets the preset conditions, then stop updating the model parameters to obtain the trained fixed filter selection model.

[0020] In a possible implementation, the target loss function is a cross entropy loss function.

[0021] In a possible implementation, the convolutional layer is a convolutional neural network (CNN).

[0022] In a possible implementation, the anti-phase noise control signal is a digital signal; before playing the anti-phase noise control signal through the in-vehicle speaker of the target vehicle so that the target road noise signal and the anti-phase noise control signal interfere with each other and cancel each other, and realizing active noise reduction for the target vehicle, the method further includes:

[0023] Converting the anti-phase noise control signal into an analog signal;

[0024] The method of playing the anti-phase noise control signal through the in-vehicle speaker of the target vehicle so that the target road noise signal and the anti-phase noise control signal interfere with each other and cancel each other out, thereby achieving active noise reduction for the target vehicle, includes:

[0025] The analog signal is played through the in-vehicle speaker of the target vehicle so that the target road noise signal and the analog signal interfere with each other and cancel each other, thereby achieving active noise reduction for the target vehicle.

[0026] In a possible implementation, the fixed filter is a fixed finite impulse response (FIR) filter.

[0027] The embodiment of the present application also provides a vehicle noise reduction device, comprising:

[0028] A first acquisition unit is used to acquire a target road noise signal during the driving process of the target vehicle;

[0029] A prediction unit is used to input the target road noise signal into a pre-built fixed filter selection model to predict a target fixed filter for noise reduction of the target vehicle; wherein the target fixed filter is determined from N fixed filters pre-installed inside the target vehicle; the parameters of the N fixed filters are determined by performing optimal filter calculation on the road noise signals under N typical working conditions collected in advance; the fixed filter selection model is obtained by jointly training the road noise signals under N working conditions collected in advance and the N fixed filters; and N is a positive integer greater than 0;

[0030] The noise reduction unit is used to filter the target road noise signal using the target fixed filter to obtain an anti-phase noise control signal; and play the anti-phase noise control signal through the in-vehicle speaker of the target vehicle so that the target road noise signal and the anti-phase noise control signal interfere with each other to achieve active noise reduction for the target vehicle.

[0031] In a possible implementation manner, the device further includes:

[0032] A second acquisition unit is used to acquire vibration and acceleration signals generated when the sample vehicle is traveling under N typical working conditions using an acceleration sensor pre-installed in the sample vehicle; and to acquire in-vehicle noise signals generated when the sample vehicle is traveling under N typical working conditions using a microphone pre-installed in the sample vehicle; and to form a sample road noise signal using the vibration and acceleration signals and the in-vehicle noise signal;

[0033] a cleaning unit, configured to clean the sample road noise signal to obtain a cleaned sample road noise signal;

[0034] The first calculation unit is used to calculate the optimal filter for the cleaned sample road noise signal by using the Wiener filter calculation method, so as to maximize the noise reduction amount of the sample road noise signal, so as to determine the parameters of N fixed filters.

[0035] In a possible implementation manner, the device further includes:

[0036] A filtering unit, configured to filter the vibration and acceleration signals using the N fixed filters, and play the filtered signals through a speaker in the sample vehicle, so that a microphone pre-installed in the sample vehicle collects the signal played by the speaker as a sample control signal;

[0037] A second calculation unit is used to calculate the value of the residual noise by using the sample control signal and the in-vehicle noise signal; and determine the Kth fixed filter corresponding to the minimum residual noise value as the filter label corresponding to the vibration and acceleration signal; K is a positive integer greater than 0 and not greater than N;

[0038] The training unit is used to train the initial fixed filter selection model using the vibration and acceleration signals and their corresponding filter labels and target loss function to obtain the fixed filter selection model.

[0039] In a possible implementation, the initial fixed filter selection model includes two one-dimensional convolutional layers and a linear layer; and the training unit includes:

[0040] A prediction subunit, used for inputting the vibration and acceleration signals into two one-dimensional convolutional layers and a linear layer in an initial fixed filter selection model, and predicting an Mth fixed filter for denoising the sample vehicle; M is a positive integer greater than 0 and not greater than N;

[0041] The judgment subunit is used to judge whether the values ​​of M and K are consistent. If so, the fixed filter selection model is obtained by training; if not, the vibration and acceleration signals are re-selected to train the initial fixed filter selection model, and during the training process, the back propagation algorithm is used to optimize the value of the target loss function calculated using the output result of the model, so as to update the initial fixed filter selection model with the optimized value, until the optimized value meets the preset conditions, then the updating of the model parameters is stopped to obtain the trained fixed filter selection model.

[0042] In a possible implementation, the target loss function is a cross entropy loss function.

[0043] In a possible implementation, the convolutional layer is a convolutional neural network (CNN).

[0044] In a possible implementation manner, the anti-phase noise control signal is a digital signal; and the device further includes:

[0045] A conversion unit, used for converting the inverted noise control signal into an analog signal;

[0046] The noise reduction unit is specifically used for:

[0047] The analog signal is played through the in-vehicle speaker of the target vehicle so that the target road noise signal and the analog signal interfere with each other and cancel each other, thereby achieving active noise reduction for the target vehicle.

[0048] In a possible implementation, the fixed filter is a fixed finite impulse response (FIR) filter.

[0049] The embodiment of the present application also provides a vehicle noise reduction device, including: a processor, a memory, and a system bus;

[0050] The processor and the memory are connected via the system bus;

[0051] The memory is used to store one or more programs, and the one or more programs include instructions. When the instructions are executed by the processor, the processor executes any one of the implementations of the above-mentioned vehicle noise reduction method.

[0052] An embodiment of the present application further provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes any one of the implementations of the above-mentioned vehicle noise reduction method.

[0053] The embodiment of the present application further provides a computer program product. When the computer program product is run on a terminal device, the terminal device executes any one of the implementations of the above-mentioned vehicle noise reduction method.

[0054] A vehicle noise reduction method, device, storage medium and equipment provided by an embodiment of the present application first obtain a target road noise signal during the driving process of a target vehicle, and then input the target road noise signal into a pre-constructed fixed filter selection model to predict a target fixed filter for noise reduction of the target vehicle; wherein the target fixed filter is determined from N fixed filters pre-installed inside the target vehicle; the parameters of the N fixed filters are determined by performing optimal filter calculation on the road noise signals under N typical working conditions collected in advance; the fixed filter selection model is obtained by training the road noise signals under N working conditions collected in advance and the N fixed filters; N is a positive integer greater than 0; then, the target road noise signal can be filtered by using the target fixed filter to obtain an anti-phase noise control signal; and the anti-phase noise control signal is played through the in-vehicle speaker of the target vehicle, so that the target road noise signal and the anti-phase noise control signal interfere with each other to achieve active noise reduction of the target vehicle.

[0055] It can be seen that since the present application first performs an optimal filter calculation on the road noise signals under N typical working conditions collected in advance, and then determines the parameters of N fixed filters with the best noise reduction effect, and then uses these N fixed filters and the road noise signal to construct a fixed filter selection model, the fixed filter selection model can be used to select the most suitable target fixed filter, and the target road noise signal during the driving process of the target vehicle can be filtered, and then the anti-phase noise control signal obtained after filtering can be used to interfere with the target road noise signal to achieve active noise reduction of the target vehicle. In this way, the noise reduction effect can be improved, the selection cost of the filter can be reduced, and the application range is wider. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0057] Figure 1 A schematic diagram of a process flow of a vehicle noise reduction method provided in an embodiment of the present application;

[0058] Figure 2 An example diagram of the deployment positions of the acceleration sensor and the speaker on the target vehicle provided in the embodiment of the present application;

[0059] Figure 3 A schematic diagram of a process of inputting a target road noise signal into a fixed filter selection model to predict a target fixed filter provided by an embodiment of the present application;

[0060] Figure 4 An example diagram of the deployment positions of the acceleration sensor, microphone and speaker on the sample vehicle provided in the embodiment of the present application;

[0061] Figure 5 A schematic diagram of a process for constructing a fixed filter selection model provided in an embodiment of the present application;

[0062] Figure 6 A schematic diagram of the composition of a vehicle noise reduction device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0063] With the rapid development of the automotive industry, people have higher and higher requirements for the NVH (Noise, Vibration, Harshness) of automobiles. For example, people may relieve stress or refresh themselves by playing music while riding in a car. However, vehicle road noise may affect the user's perception of the music in the car or other sounds outside the car, reducing the user's riding experience. Therefore, how to reduce or eliminate vehicle road noise is an increasingly important design consideration for vehicle manufacturers.

[0064] Among them, the process of vehicle road noise generation is: the vehicle tire generates random broadband vibrations at the road contact point, and is transmitted into the car body through structural transmission and air. The structural transmission part includes vibrations transmitted to the car body through the suspension system, and then radiates sound into the car body. It is the main source of interior noise below 500 Hz. The low-frequency characteristics of structure-borne road noise make it difficult to control through passive technologies commonly used in automobiles, such as viscoelastic vibration reduction treatment applied to body panels or porous materials in the car body. These methods have poor noise reduction effects at low frequencies because the wavelength of sound and vibration is similar to the thickness of the material. Therefore, at present, the noise reduction method for road noise is usually achieved by adaptive filtering algorithms, which are more effective at low frequencies and are suitable for low-frequency structure-borne road noise in cars. However, traditional linear adaptive filtering algorithms such as the adaptive filtering algorithm based on the minimum mean square error criterion (Filtered-X Least Mean Square, FxLMS) commonly used in the industry have the following disadvantages:

[0065] 1. Prone to divergence, especially when there is high variability in the secondary path or the output is saturated due to excessive noise levels. This may lead to ineffective noise cancellation and potential instability of the noise reduction system.

[0066] 2. Slow convergence. Adaptive active noise control (ANC) algorithms may converge slowly, which limits their ability to respond quickly to dynamic or rapidly changing noise environments. This slow response may result in reduced noise reduction when the surrounding noise fluctuates greatly.

[0067] 3. It is difficult to achieve noise reduction for complex scenes. This is because for complex scenes, the algorithm needs to be calibrated and adjusted to adapt to various noise conditions, so the implementation of the adaptive ANC algorithm may be complicated. Compared with the method of selecting a suitable fixed filter for vehicle noise reduction proposed in this application (which will be described in detail in subsequent embodiments), this complexity makes deployment more challenging, because the method of selecting a suitable fixed filter for vehicle noise reduction provided by this application is simpler to deploy and more robust to implement in actual environments.

[0068] 4. Large amount of calculation. Since the adaptive ANC algorithm requires real-time calculation, it requires a large amount of calculation. Compared with the method proposed in this application of selecting a suitable fixed filter for vehicle noise reduction, this large amount of calculation makes it more difficult to deploy on a platform with low computing power. This increases the computational load and hardware cost of the noise reduction system.

[0069] Therefore, the current noise reduction method for vehicle road noise not only has a high noise reduction cost, but also has a poor noise reduction effect.

[0070] To solve the above defects, the present application provides a vehicle noise reduction method, which first obtains a target road noise signal during the driving process of the target vehicle, and then inputs the target road noise signal into a pre-constructed fixed filter selection model to predict a target fixed filter for noise reduction of the target vehicle; wherein the target fixed filter is determined from N fixed filters pre-installed inside the target vehicle; the parameters of the N fixed filters are determined by performing optimal filter calculation on the road noise signals under N typical working conditions collected in advance; the fixed filter selection model is obtained by jointly training the road noise signals under N working conditions collected in advance (such as N different working conditions in which the vehicle travels at different speeds on roads of different smoothness) and N fixed filters; N is a positive integer greater than 0; then, the target road noise signal can be filtered by using the target fixed filter to obtain an anti-phase noise control signal; and the anti-phase noise control signal is played through the in-vehicle speaker of the target vehicle, so that the target road noise signal and the anti-phase noise control signal interfere with each other to achieve active noise reduction of the target vehicle.

[0071] It can be seen that since the present application first performs an optimal filter calculation on the road noise signals under N typical working conditions collected in advance, and then determines the parameters of N fixed filters with the best noise reduction effect, and then uses these N fixed filters and the road noise signal to construct a fixed filter selection model, the fixed filter selection model can be used to select the most suitable target fixed filter, and the target road noise signal during the driving process of the target vehicle can be filtered, and then the anti-phase noise control signal obtained after filtering can be used to interfere with the target road noise signal to achieve active noise reduction of the target vehicle. In this way, the noise reduction effect can be improved, the selection cost of the filter can be reduced, and the application range is wider.

[0072] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are 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.

[0073] First embodiment

[0074] See also Figure 1 , is a flow chart of a vehicle noise reduction method provided in this embodiment, the method comprising the following steps:

[0075] S101: Acquire a target road noise signal during the driving process of a target vehicle.

[0076] In this embodiment, any vehicle subjected to noise reduction processing by this embodiment is defined as a target vehicle. In order to improve the noise reduction effect on the target vehicle and improve the riding experience and riding safety of the vehicle users, firstly, it is necessary to use various acceleration sensors (also called accelerometers) pre-installed on the target vehicle to obtain the vibration and acceleration data generated by the target vehicle during the driving process as the target road noise signal to perform the subsequent step S102.

[0077] It should be noted that the present application does not limit the specific structure, quantity, and installation position of the acceleration sensor pre-installed on the target vehicle. The acceleration sensor may be a three-axis acceleration sensor, or may be replaced by other sensors that can obtain road noise signals, and the number may be one or more (such as Figure 2 4 are shown), the setting position can be as follows Figure 2 The target vehicle is shown as being located at the bottom (or other location) of the target vehicle. In addition, the target vehicle is also pre-configured with an in-vehicle speaker to play the anti-phase noise control signal determined by the subsequent steps S102-S103 to achieve active noise reduction for the target vehicle. The present application does not limit the specific structure, number, and installation location of the in-vehicle speaker on the target vehicle. The number can be one or more (such as Figure 2 6 are shown), the setting position can be as follows Figure 2 location shown or other location.

[0078] It should also be noted that the present application does not limit the specific content and dimension of the target road noise signal. For example, the target road noise signal can be 12 channels of 1-second vibration and acceleration data signals obtained at a sampling rate of 16k through 4 three-axis acceleration sensors pre-installed on the target vehicle. The dimension of the target road noise signal is 16000*12.

[0079] S102: Input the target road noise signal into a pre-built fixed filter selection model to predict a target fixed filter for noise reduction of the target vehicle; wherein the target fixed filter is determined from N fixed filters pre-installed inside the target vehicle; the parameters of the N fixed filters are determined by performing optimal filter calculation on the road noise signals under N typical working conditions collected in advance; the fixed filter selection model is obtained by jointly training the road noise signals under N working conditions collected in advance and the N fixed filters; N is a positive integer greater than 0.

[0080] In this embodiment, after the target road noise signal during the driving process of the target vehicle is obtained in step S101, in order to effectively improve the noise reduction effect on the target vehicle and reduce the noise reduction cost, the target road noise signal can be further input into a pre-constructed fixed filter selection model, so as to predict and obtain a fixed filter for noise reduction on the target vehicle, and define it as a target fixed filter to execute the subsequent step S103.

[0081] It should be noted that, in order to improve the noise reduction effect of the vehicle and reduce the noise reduction cost, the present application proposes a method for vehicle noise reduction by selecting a suitable fixed filter. Specifically, first, the road noise signal generated by the vehicle under N (N is a positive integer greater than 0) typical working conditions is collected by an acceleration sensor, and then, after performing optimal filter calculation on these road noise signals, the parameters of N fixed filters (the specific composition structure is not limited, and a fixed finite impulse response (FIR) filter can be used) that can achieve the maximum noise reduction are determined. Then, using these N fixed filters and the road noise signals under N working conditions collected in advance, a fixed filter selection model is constructed, such as Figure 3 As shown in Figure 2, the model includes convolutional layers and linear layers (such as Figure 3 In this way, after N fixed filters are installed inside the target vehicle, the target road noise signal can be input into the fixed filter selection model, so that the target fixed filter with the maximum noise reduction for the target vehicle can be selected from the N fixed filters to achieve effective noise reduction for the target vehicle.

[0082] It should be noted that, in order to improve the noise reduction effect and select a more suitable fixed filter to reduce the noise of the target vehicle, as many road noise signals generated by sample vehicles (without limiting the vehicle composition, only as a distinction from the target vehicle) under various working conditions as possible can be collected in advance. However, considering the signal collection cost and calculation amount, the road noise signals generated by the vehicle under N typical working conditions can be collected to calculate the optimal filter and then determine the parameters of the N fixed filters that can maximize the noise reduction.

[0083] Furthermore, in order to improve the accuracy of determining the parameters of the fixed filter, the present application not only installs an acceleration sensor and an in-car speaker on the sample vehicle, but also installs an in-car microphone. In this way, when the driver of the sample vehicle drives on different road surfaces (road surface types may include but are not limited to cement road surfaces, asphalt road surfaces, brick road surfaces, crushed stone road surfaces, sandstone road surfaces, gravel road surfaces, other special structured road surfaces, etc.) and different speeds (40km / h, 60km / h, 80km / h, 100km / h, etc.) under N typical working conditions, the acceleration sensor pre-installed in the sample vehicle can be used to obtain the vibration and acceleration signals generated when the sample vehicle is driving under these N typical working conditions; and the in-car microphone pre-installed in the sample vehicle can be used to obtain the in-car noise signals generated when the sample vehicle is driving under these N typical working conditions, so as to determine the parameters of the N fixed filters.

[0084] The present application does not limit the specific structure, quantity, and installation location of the acceleration sensor and the in-car speaker pre-installed in the sample vehicle. Please refer to the relevant introduction of the acceleration sensor and the in-car speaker pre-installed in the target vehicle in step S101. In addition, the present application does not limit the specific structure, quantity, and installation location of the in-car microphone pre-installed in the sample vehicle. The quantity can be one or more (such as Figure 4 4 are shown), the setting position can be as follows Figure 4 As shown, they are located at the main driver's seat, the co-driver's seat, the left rear seat, the right rear seat or other positions. It is understandable that, depending on the model of the sample vehicle (or target vehicle), the acceleration sensor, the speaker and the microphone can be arranged in a variety of different ways, which will not be described one by one here.

[0085] Specifically, an optional implementation method is that the process of determining the parameters of these N fixed filters can be: first, using the acceleration sensor pre-installed in the sample vehicle to obtain the vibration and acceleration signals generated by the sample vehicle when it is driving under N typical working conditions; and using the microphone pre-installed in the sample vehicle to obtain the in-vehicle noise signal generated by the sample vehicle when it is driving under N typical working conditions, and using the vibration and acceleration signals and the in-vehicle noise signal to form a sample road noise signal.

[0086] Among them, the present application does not limit the specific content of the N typical working conditions, and only needs to ensure that the sample road noise signals obtained can cover different typical vehicle conditions and road conditions as much as possible. In addition, the present application does not limit the specific number of sample road noise signals. For example, hundreds of hours of data signal collection can be performed to ensure that the collected training data can more accurately determine the parameters of the N fixed filters, and can achieve better training results in the subsequent training process of the fixed filter selection model.

[0087] Then, the collected sample road noise signals (including vibration and acceleration signals generated by the sample vehicle and in-vehicle noise signals) can be cleaned using existing or future data cleaning methods to obtain cleaned sample road noise signals. For example, data cleaning can be performed manually or by machine to remove data that does not meet the conditions in the sample road noise signals, such as signal data generated when the sample vehicle is idling or signal data with a volume that is too loud and exceeds the range (such as the signal range collected by the microphone is 0~1, etc.). The specific cleaning method can be: by monitoring parameters such as the speed or engine speed of the sample vehicle, to identify and remove the idle sections in the sample road noise signal data (such as when the speed or engine speed of the sample vehicle is 0, the idle data is generated), and only retain data related to the actual driving situation. In the non-idling sections, it is also necessary to detect and remove any abnormal values ​​or unreasonable data points to obtain the cleaned sample road noise signal.

[0088] Next, the Wiener filter calculation method may be used to calculate the optimal filter for the cleaned sample road noise signal, so that the noise reduction amount of the sample road noise signal is maximized, thereby determining the parameters of the N fixed filters.

[0089] Among them, the specific method of calculating the optimal filter for the cleaned sample road noise signal using the Wiener filter calculation method is: using the cross power spectrum of the noise at the microphone and the acceleration data divided by the auto power spectrum of the acceleration data, the parameters of N offline fixed filters (such as fixed FIR filters) can be determined according to the calculation results. The form of these N fixed filters can be the same, for example, they can all be fixed FIR filters, but the parameters of each fixed filter are different. For example, each fixed FIR filter can have 128 parameters. If these N fixed filters are all fixed FIR filters, at least one parameter of the 128 parameters of each fixed filter is different. Then, these N offline fixed filters (such as fixed FIR filters) with determined parameters can be installed inside the target vehicle (such as stored in memory), and the fixed filter selection model can select the most suitable target fixed filter for noise reduction of the target vehicle according to the target road noise signal during the driving process of the target vehicle, so that the noise reduction amount is maximized.

[0090] Next, this embodiment will introduce the construction process of the fixed filter selection model. Figure 5 As shown, it shows a schematic diagram of the process of constructing a fixed filter selection model provided by this embodiment. The specific construction process of the fixed filter selection model may include the following steps AC:

[0091] Step A: Filter the vibration and acceleration signals using N fixed filters, and play the filtered signals through the speakers in the sample vehicle, so that the microphone pre-installed in the sample vehicle can collect the signals played by the speakers as sample control signals.

[0092] In this embodiment, in order to construct a fixed filter selection model, a lot of preparatory work needs to be done in advance. First, it is necessary to use the acceleration sensor pre-installed in the sample vehicle to obtain the vibration and acceleration signals generated when the sample vehicle is driving under N typical working conditions; and use the microphone pre-installed in the sample vehicle to obtain the in-vehicle noise signal (represented by y) generated when the sample vehicle is driving under N typical working conditions. These signal data can cover as many vehicle driving conditions as possible (such as sample vehicles driving at different speeds on different types of road conditions).

[0093] Then, the vibration and acceleration signals generated by the sample vehicle under N typical working conditions can be filtered using N fixed filters determined by the Wiener filter calculation method (the specific determination process will not be repeated here) to obtain the filtered signals. The filtered signals are played through the speakers in the sample vehicle. In this way, the impulse response from the speaker to the microphone can be convolved with the filtered signal to obtain the signal propagated to the microphone (that is, the signal played by the speaker collected by the microphone pre-installed in the sample vehicle), and it is defined as the sample control signal. Indicates that it is used to execute the subsequent step B.

[0094] Step B: Calculate the value of the residual noise using the sample control signal and the in-vehicle noise signal; and determine the Kth fixed filter corresponding to the minimum residual noise value as the filter label corresponding to the vibration and acceleration signals; wherein K is a positive integer greater than 0 and not greater than N.

[0095] In this embodiment, in order to construct a fixed filter selection model, after obtaining the in-vehicle noise signal y and the sample control signal in step A, After that, the value of the residual noise (represented by L) is further calculated using the values ​​of the two, so as to determine the fixed filter with the best noise reduction effect from the N fixed filters according to the value results, and define it as the Kth (where K is a positive integer greater than 0 and not greater than N) fixed filter, that is, the fixed filter corresponding to the minimum value of the residual noise L (that is, the maximum noise reduction amount) is defined as the Kth fixed filter, which is used as the filter label corresponding to the vibration and acceleration signals generated by the sample vehicle at the same time, so as to execute the subsequent step C.

[0096] The calculation formula of the residual noise L is as follows:

[0097]

[0098] Wherein, L represents the residual noise. It can be understood that the smaller the value of the residual noise L is, the greater the noise reduction amount and the better the noise reduction effect; y represents the in-vehicle noise signal collected by the in-vehicle microphone pre-installed in the sample vehicle; represents the signal played by the speaker picked up by the microphone pre-installed in the sample vehicle; Indicates the parameters of the corresponding fixed filter, and the specific value is not limited.

[0099] Step C: Using the vibration and acceleration signals and their corresponding filter labels and target loss function, the initial fixed filter selection model is trained to obtain a fixed filter selection model.

[0100] It should be noted that the initial fixed filter selection model used in this application includes two one-dimensional convolutional layers and a linear layer. The specific composition of the convolutional layer and the linear layer is not limited and can be set according to actual conditions and experience. For example, the two one-dimensional convolutional layers can be set as a convolutional neural network (CNN). Figure 5 shown.

[0101] Thus, in this embodiment, after the Kth fixed filter is used as the filter label corresponding to the vibration and acceleration signals in step B, the vibration and acceleration signals (such as 12 channels of 1-second vibration and acceleration data signals obtained by 4 triaxial acceleration sensors pre-installed on the sample vehicle, with a dimension of 16000*12) can be further used, combined with the target loss function (the specific function content is not limited and can be selected according to actual conditions and experience values, such as it can be selected as a cross entropy loss function), and the corresponding filter label (i.e., the Kth fixed filter) to train the initial fixed filter selection model.

[0102] Moreover, during the training process, the vibration and acceleration signals (such as 12 channels of 1 second vibration and acceleration data signals obtained by 4 triaxial acceleration sensors pre-installed on the sample vehicle, with a dimension of 16000*12) can first be input into the two one-dimensional convolutional layers and linear layers in the initial fixed filter selection model for selection, and the model outputs a selection probability distribution (i.e., a set of vectors) for N fixed filters. The value of each dimension in the vector distribution can be a value in the interval [0,1], and the value of each dimension represents the probability value of the selected fixed filter. At this time, the fixed filter corresponding to the maximum probability value can be used as the predicted fixed filter for denoising the sample vehicle, so that the most suitable fixed filter for denoising the sample vehicle (i.e., the largest amount of denoising) can be selected from the N fixed filters, and it is defined as the Mth fixed filter, where M is a positive integer greater than 0 and not greater than N.

[0103] For example: assuming that the output vector of the model is s=[0.90, 0.01, 0.02, 0.02, 0.01, 0.01, 0.03…], and the vector s has N dimensions, where the dimension values ​​represented by the ellipsis are all less than 0.90, it can be seen that the value of the first dimension, 0.90, is the highest. Therefore, the first fixed filter corresponding to this dimension is the most suitable fixed filter for denoising the sample vehicle (i.e., the largest amount of denoising), that is, M takes the value of 1.

[0104] Then, determine whether the values ​​of M and K are consistent. If so, the model prediction is correct and the fixed filter selection model is obtained through training. If not, the model prediction is wrong, and it is necessary to reselect vibration and acceleration signals to train the initial fixed filter selection model. During the training process, the gradient back propagation algorithm is used to optimize the value of the target loss function calculated using the output result of the model, so as to use the optimized value to update the model parameters for training until the optimized value of the target loss function meets the preset conditions, such as reaching the minimum value with a small change (basically unchanged), or reaching the preset maximum number of iterations (such as 100 times), then the update of the model parameters is stopped to obtain the trained fixed filter selection model.

[0105] It should be noted that after each round of training, the validation set can also be used to evaluate the model to avoid overfitting. In this way, the model selection and parameter tuning using the target loss function (such as the cross entropy loss function) can improve the accuracy and generalization ability of the model. After multiple rounds of training and parameter tuning, the model with the smallest target loss function (such as the cross entropy loss function) in the test set can be selected as the final fixed filter selection model.

[0106] S103: Filtering the target road noise signal with a target fixed filter to obtain an anti-phase noise control signal; and playing the anti-phase noise control signal through a speaker in the target vehicle, so that the target road noise signal and the anti-phase noise control signal interfere with each other and cancel each other, thereby achieving active noise reduction for the target vehicle.

[0107] In this embodiment, after the target fixed filter for noise reduction of the target vehicle is predicted in step S102, the target fixed filter can be further used to filter the target road noise signal obtained in step S101 to obtain a filtered signal, and an anti-phase noise control signal is defined as the anti-phase noise control signal. The amplitude of the anti-phase noise control signal is the same as the target road noise signal, but the phases of the two are opposite. In this way, the anti-phase noise control signal is played through the in-vehicle speaker of the target vehicle, so that the target road noise signal and the anti-phase noise control signal can interfere with each other, reduce the amplitude of the target road noise signal, or even completely eliminate it, thereby achieving active noise reduction of the target vehicle.

[0108] Among them, an optional implementation method is that the anti-phase noise control signal is a digital signal. Before playing the anti-phase noise control signal through the in-vehicle speaker of the target vehicle, it is necessary to use an existing or future digital-to-analog conversion method to convert the digital signal of the anti-phase noise control signal into an analog signal, and then play the analog signal through the in-vehicle speaker, so that the target road noise signal and the analog signal interfere with each other, thereby achieving active noise reduction for the target vehicle.

[0109] However, it should be noted that since the specifications and parameters of each speaker are different, the speaker needs to be tuned in advance to avoid affecting the noise reduction effect. When tuning the speaker, you need to first determine the speaker's frequency response characteristics and sensitivity and other parameters, and then use the corresponding acoustic test instrument to test and calibrate the speaker. During the calibration process, you can use indicators such as frequency response flatness, phase response, distortion, sensitivity, etc. to evaluate the performance of the speaker and make adjustments to achieve the best effect.

[0110] In this way, by executing the above steps S101-S103, N fixed filters are determined by first performing offline optimal filter calculation on the pre-collected road noise signals under N typical working conditions, and then using the pre-built fixed filter selection model (offline pre-training method), according to the target road noise signal acquired in real time, the most suitable fixed filter is selected from the N fixed filters to reduce the noise of the target vehicle. This can reduce the complexity of noise reduction, and can also use low-cost hardware equipment (such as fixed filters) to reduce the cost of noise reduction.

[0111] In summary, the vehicle noise reduction method provided by the present embodiment first obtains a target road noise signal during the driving process of the target vehicle, and then inputs the target road noise signal into a pre-constructed fixed filter selection model to predict a target fixed filter for noise reduction of the target vehicle; wherein the target fixed filter is determined from N fixed filters pre-installed inside the target vehicle; the parameters of the N fixed filters are determined by performing optimal filter calculation on the road noise signals under N typical working conditions collected in advance; the fixed filter selection model is obtained by jointly training the road noise signals under N working conditions collected in advance and the N fixed filters; N is a positive integer greater than 0; then, the target road noise signal can be filtered by using the target fixed filter to obtain an anti-phase noise control signal; and the anti-phase noise control signal is played through the in-vehicle speaker of the target vehicle, so that the target road noise signal and the anti-phase noise control signal interfere with each other to achieve active noise reduction of the target vehicle.

[0112] It can be seen that since the present application first performs an optimal filter calculation on the road noise signals under N typical working conditions collected in advance, and then determines the parameters of N fixed filters with the best noise reduction effect, and then uses these N fixed filters and the road noise signal to construct a fixed filter selection model, the fixed filter selection model can be used to select the most suitable target fixed filter, and the target road noise signal during the driving process of the target vehicle can be filtered, and then the anti-phase noise control signal obtained after filtering can be used to interfere with the target road noise signal to achieve active noise reduction of the target vehicle. In this way, the noise reduction effect can be improved, the selection cost of the filter can be reduced, and the application range is wider.

[0113] Second embodiment

[0114] This embodiment will introduce a vehicle noise reduction device. For related content, please refer to the above method embodiment.

[0115] See also Figure 6 , is a schematic diagram of the composition of a vehicle noise reduction device provided in this embodiment, the device 600 includes:

[0116] The first acquisition unit 601 is used to acquire a target road noise signal during the driving process of the target vehicle;

[0117] The prediction unit 602 is used to input the target road noise signal into a pre-built fixed filter selection model to predict a target fixed filter for noise reduction of the target vehicle; wherein the target fixed filter is determined from N fixed filters pre-installed inside the target vehicle; the parameters of the N fixed filters are determined by performing optimal filter calculation on the road noise signals under N typical working conditions collected in advance; the fixed filter selection model is obtained by jointly training the road noise signals under N working conditions collected in advance and the N fixed filters; and N is a positive integer greater than 0;

[0118] The noise reduction unit 603 is used to filter the target road noise signal using the target fixed filter to obtain an anti-phase noise control signal; and play the anti-phase noise control signal through the in-vehicle speaker of the target vehicle so that the target road noise signal and the anti-phase noise control signal interfere with each other to achieve active noise reduction for the target vehicle.

[0119] In an implementation of this embodiment, the device further includes:

[0120] A second acquisition unit is used to acquire vibration and acceleration signals generated when the sample vehicle is traveling under N typical working conditions using an acceleration sensor pre-installed in the sample vehicle; and to acquire in-vehicle noise signals generated when the sample vehicle is traveling under N typical working conditions using a microphone pre-installed in the sample vehicle; and to form a sample road noise signal using the vibration and acceleration signals and the in-vehicle noise signal;

[0121] a cleaning unit, configured to clean the sample road noise signal to obtain a cleaned sample road noise signal;

[0122] The first calculation unit is used to calculate the optimal filter for the cleaned sample road noise signal by using the Wiener filter calculation method, so as to maximize the noise reduction amount of the sample road noise signal, so as to determine the parameters of N fixed filters.

[0123] In an implementation of this embodiment, the device further includes:

[0124] A filtering unit, configured to filter the vibration and acceleration signals using the N fixed filters, and play the filtered signals through a speaker in the sample vehicle, so that a microphone pre-installed in the sample vehicle collects the signal played by the speaker as a sample control signal;

[0125] A second calculation unit is used to calculate the value of the residual noise by using the sample control signal and the in-vehicle noise signal; and determine the Kth fixed filter corresponding to the minimum residual noise value as the filter label corresponding to the vibration and acceleration signal; K is a positive integer greater than 0 and not greater than N;

[0126] The training unit is used to train the initial fixed filter selection model using the vibration and acceleration signals and their corresponding filter labels and target loss function to obtain the fixed filter selection model.

[0127] In an implementation of this embodiment, the initial fixed filter selection model includes two one-dimensional convolutional layers and a linear layer; the training unit includes:

[0128] A prediction subunit, used for inputting the vibration and acceleration signals into two one-dimensional convolutional layers and a linear layer in an initial fixed filter selection model, and predicting an Mth fixed filter for denoising the sample vehicle; M is a positive integer greater than 0 and not greater than N;

[0129] The judgment subunit is used to judge whether the values ​​of M and K are consistent. If so, the fixed filter selection model is obtained by training; if not, the vibration and acceleration signals are re-selected to train the initial fixed filter selection model, and during the training process, the back propagation algorithm is used to optimize the value of the target loss function calculated using the output result of the model, so as to update the initial fixed filter selection model with the optimized value, until the optimized value meets the preset conditions, then the updating of the model parameters is stopped to obtain the trained fixed filter selection model.

[0130] In one implementation of this embodiment, the target loss function is a cross entropy loss function.

[0131] In one implementation of this embodiment, the convolution layer is a convolutional neural network (CNN).

[0132] In an implementation of this embodiment, the anti-phase noise control signal is a digital signal; the device further includes:

[0133] A conversion unit, used for converting the inverted noise control signal into an analog signal;

[0134] The noise reduction unit 603 is specifically used for:

[0135] The analog signal is played through the in-vehicle speaker of the target vehicle so that the target road noise signal and the analog signal interfere with each other and cancel each other, thereby achieving active noise reduction for the target vehicle.

[0136] In an implementation manner of this embodiment, the fixed filter is a fixed finite impulse response (FIR) filter.

[0137] Furthermore, an embodiment of the present application also provides a vehicle noise reduction device, including: a processor, a memory, and a system bus;

[0138] The processor and the memory are connected via the system bus;

[0139] The memory is used to store one or more programs, and the one or more programs include instructions. When the instructions are executed by the processor, the processor executes any one of the implementation methods of the above-mentioned vehicle noise reduction method.

[0140] Furthermore, an embodiment of the present application also provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes any one of the implementation methods of the above-mentioned vehicle noise reduction method.

[0141] Furthermore, an embodiment of the present application also provides a computer program product, which, when executed on a terminal device, enables the terminal device to execute any one of the implementation methods of the above-mentioned vehicle noise reduction method.

[0142] It can be known from the description of the above implementation mode that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., including several instructions for enabling a computer device (which can be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0143] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0144] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0145] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A vehicle noise reduction method, characterized in that: include: Acquire a target road noise signal during the driving process of the target vehicle; The target road noise signal is input into a pre-built fixed filter selection model to predict a target fixed filter for reducing noise on the target vehicle; wherein the target fixed filter is determined from N fixed filters pre-installed inside the target vehicle; the parameters of the N fixed filters are determined by performing optimal filter calculation on the road noise signals under N typical working conditions collected in advance; the fixed filter selection model is obtained by jointly training the road noise signals under N working conditions collected in advance and the N fixed filters; and N is a positive integer greater than 0; The target road noise signal is filtered by using the target fixed filter to obtain an anti-phase noise control signal; and the anti-phase noise control signal is played through a speaker in the target vehicle, so that the target road noise signal and the anti-phase noise control signal interfere with each other to achieve active noise reduction for the target vehicle; The parameters of the N fixed filters are determined as follows: Using an acceleration sensor pre-installed in the sample vehicle to obtain vibration and acceleration signals generated when the sample vehicle is driving under N typical working conditions; and using a microphone pre-installed in the sample vehicle to obtain a vehicle interior noise signal generated when the sample vehicle is driving under N typical working conditions; and using the vibration and acceleration signals and the vehicle interior noise signal to form a sample road noise signal; Cleaning the sample road noise signal to obtain a cleaned sample road noise signal; Using the Wiener filter calculation method, the optimal filter is calculated for the cleaned sample road noise signal, so that the noise reduction amount of the sample road noise signal is maximized, so as to determine the parameters of N fixed filters; The fixed filter selection model is constructed as follows: Filtering the vibration and acceleration signals by using the N fixed filters, and playing the filtered signals through a speaker in a sample vehicle, so that a microphone pre-installed in the sample vehicle collects the signal played by the speaker as a sample control signal; Calculate the value of residual noise by using the sample control signal and the in-vehicle noise signal; and determine the Kth fixed filter corresponding to the minimum residual noise value as the filter label corresponding to the vibration and acceleration signals; K is a positive integer greater than 0 and not greater than N; The initial fixed filter selection model is trained using the vibration and acceleration signals and their corresponding filter labels and target loss function to obtain the fixed filter selection model.

2. The method according to claim 1, characterized in that The initial fixed filter selection model includes two one-dimensional convolutional layers and a linear layer; the initial fixed filter selection model is trained using the vibration and acceleration signals and their corresponding filter labels and target loss functions to obtain the fixed filter selection model, including: Inputting the vibration and acceleration signals into two one-dimensional convolutional layers and a linear layer in an initial fixed filter selection model, predicting an Mth fixed filter for denoising the sample vehicle; M is a positive integer greater than 0 and not greater than N; Determine whether the values ​​of M and K are consistent. If so, train to obtain the fixed filter selection model; if not, reselect the vibration and acceleration signals to train the initial fixed filter selection model, and during the training process, use the back propagation algorithm to optimize the value of the target loss function calculated using the output result of the model, so as to use the optimized value to update the initial fixed filter selection model for training, until the optimized value meets the preset conditions, then stop updating the model parameters to obtain the trained fixed filter selection model.

3. The method according to claim 1, characterized in that: The objective loss function is a cross entropy loss function.

4. The method according to claim 2, characterized in that: The convolutional layer is a convolutional neural network (CNN).

5. The method according to claim 1, characterized in that The anti-phase noise control signal is a digital signal; before playing the anti-phase noise control signal through the in-vehicle speaker of the target vehicle so that the target road noise signal and the anti-phase noise control signal interfere with each other and cancel each other, and realizing active noise reduction for the target vehicle, the method further includes: Converting the anti-phase noise control signal into an analog signal; The method of playing the anti-phase noise control signal through the in-vehicle speaker of the target vehicle so that the target road noise signal and the anti-phase noise control signal interfere with each other and cancel each other out, thereby achieving active noise reduction for the target vehicle, includes: The analog signal is played through the in-vehicle speaker of the target vehicle so that the target road noise signal and the analog signal interfere with each other and cancel each other, thereby achieving active noise reduction for the target vehicle.

6. The method according to any one of claims 1 to 5, characterized in that: The fixed filter is a fixed finite impulse response (FIR) filter.

7. A vehicle noise reduction device, characterized in that: include: A first acquisition unit is used to acquire a target road noise signal during the driving process of the target vehicle; A prediction unit is used to input the target road noise signal into a pre-built fixed filter selection model to predict a target fixed filter for noise reduction of the target vehicle; wherein the target fixed filter is determined from N fixed filters pre-installed inside the target vehicle; the parameters of the N fixed filters are determined by performing optimal filter calculation on the road noise signals under N typical working conditions collected in advance; the fixed filter selection model is obtained by jointly training the road noise signals under N working conditions collected in advance and the N fixed filters; and N is a positive integer greater than 0; a noise reduction unit, configured to filter the target road noise signal using the target fixed filter to obtain an anti-phase noise control signal; and to play the anti-phase noise control signal through a speaker in the target vehicle, so that the target road noise signal and the anti-phase noise control signal interfere with each other to achieve active noise reduction for the target vehicle; The device also includes: A second acquisition unit is used to acquire vibration and acceleration signals generated when the sample vehicle is traveling under N typical working conditions using an acceleration sensor pre-installed in the sample vehicle; and to acquire in-vehicle noise signals generated when the sample vehicle is traveling under N typical working conditions using a microphone pre-installed in the sample vehicle; and to form a sample road noise signal using the vibration and acceleration signals and the in-vehicle noise signal; a cleaning unit, configured to clean the sample road noise signal to obtain a cleaned sample road noise signal; A first calculation unit is used to calculate the optimal filter for the cleaned sample road noise signal by using a Wiener filter calculation method, so that the noise reduction amount of the sample road noise signal is maximized, so as to determine the parameters of N fixed filters; The device also includes: A filtering unit, configured to filter the vibration and acceleration signals using the N fixed filters, and play the filtered signals through a speaker in the sample vehicle, so that a microphone pre-installed in the sample vehicle collects the signal played by the speaker as a sample control signal; A second calculation unit is used to calculate the value of the residual noise by using the sample control signal and the in-vehicle noise signal; and determine the Kth fixed filter corresponding to the minimum residual noise value as the filter label corresponding to the vibration and acceleration signal; K is a positive integer greater than 0 and not greater than N; The training unit is used to train the initial fixed filter selection model using the vibration and acceleration signals and their corresponding filter labels and target loss function to obtain the fixed filter selection model.

8. A vehicle noise reduction device, characterized in that: include: Processor, memory, system bus; The processor and the memory are connected via the system bus; The memory is used to store one or more programs, wherein the one or more programs include instructions, and when the instructions are executed by the processor, the processor executes the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on a terminal device, the terminal device executes the method according to any one of claims 1 to 6.

Images