In-vehicle sound quality partition evaluation method, control method, system, and storage medium

Abstract

The present application relates to a kind of in-vehicle sound quality zoning evaluation method, control method, system and storage medium, belong to the field of automobile NVH performance.In-vehicle sound quality zoning evaluation method includes: obtaining the sound information of current area in vehicle;Sound information is input in the sound quality mode corresponding to the sound quality evaluation model of current area and is evaluated, obtains the sound quality score of current area, the sound quality evaluation model is the hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm.This method will nonlinear mapping ability, and can automatically extract the convolutional neural network of image feature in sound quality evaluation and prediction modeling, combine the adaptive feature extraction capability possessed by convolutional neural network and the advantage of support vector regression algorithm in small data sample scene, establishes in-vehicle sound quality evaluation model, different sound quality mode corresponds to different sound quality evaluation model, realizes the evaluation and prediction of in-vehicle sound quality in different sound quality mode of region.

Description

Technical Field

[0001] This invention relates to the field of automotive NVH performance, specifically to a method for evaluating in-vehicle sound quality zoning, a system for evaluating in-vehicle sound quality zoning, a method for actively controlling in-vehicle sound quality zoning, a device for actively controlling in-vehicle sound quality zoning, a vehicle, and a computer-readable storage medium. Background Technology

[0002] Sound quality directly reflects the subjective perception of car sound by drivers and passengers, and is an important factor influencing customer purchasing decisions. With economic development and the improvement of people's material and cultural living standards, users' needs for in-vehicle sound quality are becoming increasingly multidimensional. Different customer groups have different needs; some prefer a quiet and comfortable in-vehicle environment, while others pursue a sporty driving experience. Even within the same in-vehicle space, the sound quality needs of occupants in different positions, such as the driver's seat and rear passenger seats, vary. Therefore, researching multidimensional sound quality evaluation and control methods for passenger vehicles is of practical significance.

[0003] Subjective sound quality evaluation methods can intuitively reflect human subjective feelings, but they rely on the evaluator's experience and work status, resulting in poor stability and hindering data accumulation for enterprises. Therefore, many scholars have attempted to establish a relationship model between subjective evaluation results and objective sound measurement parameters to achieve objectification of sound quality evaluation. CN103471709A, "Method for Predicting Sound Quality of Passenger Car Interior Noise," uses objective psychoacoustic parameters of in-vehicle sound quality, such as A-weighted sound level, loudness, and sharpness, along with subjective sound quality evaluation results, as a basis to establish an objective quantitative model for subjective in-vehicle sound quality evaluation through a BP neural network.

[0004] However, the above methods require manual extraction of objective acoustic indicators, which is not only time-consuming and labor-intensive, but also the rationality and comprehensiveness of the extracted indicators depend to a large extent on the experience and cognition of engineers or researchers, making it difficult to truly and comprehensively reflect the physical characteristics of the sound sample. Summary of the Invention

[0005] One objective of this invention is to provide a method for evaluating in-vehicle sound quality by zoning. This method utilizes convolutional neural networks (CNNs), which possess strong nonlinear mapping capabilities and can automatically extract image features, in sound quality evaluation and prediction modeling. Combining the adaptive feature extraction capabilities of CNNs with the advantages of support vector regression algorithms in small data sample scenarios, an in-vehicle sound quality evaluation model combining deep CNNs and support vector regression algorithms is established. Different sound quality modes correspond to different sound quality evaluation models, enabling the evaluation and prediction of in-vehicle sound quality by region and sound quality mode, with high model accuracy. A second objective is to provide a method for active control of in-vehicle sound quality by zoning. Based on the in-vehicle sound quality score and the target score corresponding to the sound quality mode of the current region, an active noise control system is used to actively control the noise in the current region, efficiently achieving the evaluation and improvement of in-vehicle sound quality. This allows different users to have different private acoustic environments according to their own needs within the same space, satisfying the personalized driving experience of different drivers and passengers.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] A method for evaluating in-vehicle sound quality by zone, the method comprising:

[0008] Obtain sound information for the current area inside the vehicle;

[0009] The sound information is input into the sound quality evaluation model corresponding to the sound quality mode of the current area for evaluation, and the sound quality score of the current area is obtained. The sound quality evaluation model is a hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm.

[0010] Based on the above technical means, a hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm was used to construct a sound quality evaluation model corresponding to different sound quality modes. The sound information of different areas in the vehicle is evaluated according to the sound quality mode requirements of users in different areas, and the evaluation accuracy is high.

[0011] In this embodiment of the application, the sound quality evaluation model is trained using the following method:

[0012] Collect noise samples;

[0013] The noise samples are classified according to the sound quality dimensions corresponding to different sound quality modes to obtain the noise samples corresponding to each sound quality mode.

[0014] Scoring and labeling the noise samples corresponding to each sound quality mode;

[0015] The noise samples and their corresponding scores for each sound quality mode are divided into a training set and a validation set.

[0016] The hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm was trained using training set and validation set to obtain the sound quality evaluation model corresponding to each sound quality mode.

[0017] Based on the above technical means, sound quality evaluation models corresponding to different sound qualities can be trained and used to achieve sound quality evaluation under different sound quality modes.

[0018] In this embodiment of the application, the noise samples include noise samples collected under different working conditions, different types of road surfaces, different test locations inside the vehicle, different speeds, and on different vehicles.

[0019] The noise samples collected using the above-mentioned techniques can reflect the diverse noise generated by the vehicle during actual use. Different test locations inside the vehicle can reflect the noise situation in different areas of the vehicle. Such noise samples are used for model training, and the trained model has higher accuracy.

[0020] In this embodiment, the hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and a regression output layer.

[0021] The prediction model constructed using the aforementioned techniques possesses both the adaptive feature extraction capabilities of convolutional neural network algorithms and the advantages of support vector regression algorithms in scenarios with small data samples. It is more suitable for scenarios where noise sample collection is challenging, such as in-vehicle sound quality evaluation.

[0022] A second aspect of the present invention provides an in-vehicle sound quality zoning evaluation system, the system comprising:

[0023] The data acquisition unit is used to acquire sound information of the current area inside the vehicle;

[0024] The sound quality scoring unit is used to input sound information into the sound quality evaluation model corresponding to the sound quality mode of the current area for evaluation, and obtain the sound quality score of the current area.

[0025] The above-mentioned technical means can be used to evaluate the sound information collected in the current area of ​​the vehicle. The evaluation is based on a hybrid prediction model of deep convolutional neural network algorithm and support vector regression algorithm. Different sound quality modes correspond to different sound quality evaluation models, and the evaluation accuracy is high.

[0026] A third aspect of the present invention provides a method for active control of in-vehicle sound quality zoning, the method comprising:

[0027] The sound quality score of the current area is obtained according to the in-vehicle sound quality zoning evaluation method described above.

[0028] Compare the sound quality score of the current area with the target score corresponding to the sound quality pattern of the current area;

[0029] If the sound quality score is lower than the target score, the active noise control system will actively control the noise in the current area.

[0030] Repeat the above steps until the sound quality score of the current area is not lower than the target score.

[0031] The above technical solution can achieve sound quality scoring in different areas of the vehicle. Then, based on the sound quality scores and the target score of the set sound quality mode, active noise control can be performed in the current area. This efficiently achieves the evaluation and improvement of in-vehicle sound quality, allowing different users to have different private acoustic environments according to their own needs within the same space, thus satisfying the personalized driving experience of different drivers and passengers.

[0032] In this embodiment of the application, controlling the active noise control system to actively control noise in the current area includes:

[0033] Acquire reference signals and sound information for the current area;

[0034] Calculate the secondary sound source control signal based on the reference signal and the sound information of the current area;

[0035] The secondary sound source is controlled to generate secondary sound waves according to the secondary sound source control signal.

[0036] Based on the aforementioned technical means, the secondary sound source is controlled to emit sound according to the reference signal collected from the vehicle and the sound information of the current area. The sound waves emitted by the secondary sound source interfere with and cancel out the noise in the current area, thereby improving the sound quality score of the current area.

[0037] In this embodiment of the application, calculating the secondary sound source control signal based on the reference signal and the sound information of the current area includes:

[0038] The secondary sound source control signal is calculated using an artificial neural network algorithm based on the reference signal and the sound information of the current area.

[0039] Using the above-mentioned technical means, an artificial neural network algorithm is used to calculate the control signal of the secondary sound source. The artificial neural network algorithm has strong adaptive learning ability and nonlinear characteristics, which can better calculate the weights of the control algorithm and improve the sound quality inside the vehicle.

[0040] A fourth aspect of the present invention provides an in-vehicle sound quality zoning active control device, the device comprising:

[0041] The in-vehicle sound quality zoning evaluation system is used to obtain the sound quality score of the current area;

[0042] The judgment unit is used to compare the sound quality score of the current area with the target score corresponding to the sound quality pattern of the current area;

[0043] The active control unit is used to control the active noise control system to actively control noise in the current area when the sound quality score is lower than the target score.

[0044] According to the above technical solution, the device can achieve sound quality scoring in different areas of the vehicle, and then, based on the sound quality scores and the target score of the set sound quality mode as the standard, actively control the noise in the current area, efficiently realize the evaluation and improvement of the sound quality in the vehicle, so that different users can have different private acoustic environments according to their own needs in the same space, and meet the personalized driving experience of different drivers and passengers.

[0045] In this embodiment of the application, the active noise control system includes:

[0046] Multiple vibration sensors are placed at the locations where vehicle noise is generated to collect reference signals;

[0047] Multiple secondary sound sources are arranged in different areas of the vehicle to generate secondary sound waves according to the control signals of the secondary sound sources;

[0048] Multiple error sensors are deployed in different areas of the vehicle to collect sound information from different areas inside the vehicle;

[0049] The controller is used to generate control signals for the secondary sound source.

[0050] Through the above-mentioned technical means, accurate reference signals and corresponding regional sound information can be collected, enabling targeted regional sound quality control.

[0051] The fifth aspect of the present invention provides a vehicle in which the in-vehicle sound quality zoning active control method described above is applied to perform in-vehicle sound quality zoning control.

[0052] Through the aforementioned technical means, vehicles can provide personalized driving and riding experiences for different drivers and passengers.

[0053] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the in-vehicle sound quality zoning evaluation method or the in-vehicle sound quality zoning active control method.

[0054] The beneficial effects of this invention are:

[0055] (1) This invention employs a convolutional neural network algorithm with strong nonlinear mapping capabilities and the ability to automatically extract image features. It combines the adaptive feature extraction capability of the convolutional neural network with the advantages of the support vector regression algorithm in small data sample scenarios. The model has high prediction accuracy, and the sound quality evaluation does not rely on manual extraction of objective acoustic indicators, making the sound quality evaluation more objective. Based on the sound quality mode requirements of users in different areas of the vehicle, the sound information of different areas is evaluated, providing a technical basis for meeting the needs of different drivers and passengers.

[0056] This invention can score the sound quality of different areas inside the vehicle, and then, based on the sound quality scores and the target score of the set sound quality mode, actively control the noise in the current area. This efficiently evaluates and improves the sound quality inside the vehicle, allowing different users to have different private acoustic environments according to their own needs within the same space, thus satisfying the personalized driving experience of different drivers and passengers. Attached Figure Description

[0057] Figure 1 A flowchart of a vehicle interior sound quality zoning evaluation method provided in one embodiment of the present invention;

[0058] Figure 2 A schematic diagram of the training process of a sound quality evaluation model provided in one embodiment of the present invention;

[0059] Figure 3 A schematic diagram of the acoustic quality mixing prediction model structure based on CNN-SVR provided in one embodiment of the present invention;

[0060] Figure 4 A block diagram of an in-vehicle sound quality zoning evaluation system provided in one embodiment of the present invention;

[0061] Figure 5 A flowchart of an active control method for in-vehicle sound quality zoning provided in one embodiment of the present invention;

[0062] Figure 6 A schematic diagram of active noise control provided for one embodiment of the present invention;

[0063] Figure 7 This is a schematic diagram of the interior space division provided for one embodiment of the present invention;

[0064] Figure 8 This is a schematic diagram of active process control of in-vehicle sound quality according to one embodiment of the present invention.

[0065] Among them, 100-vehicle, 2-front driver's seat area, 3-front passenger seat area, 4-rear left-side passenger area, 5-rear right-side passenger area. Detailed Implementation

[0066] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0067] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the illustrations only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0068] This invention provides a method for evaluating in-vehicle sound quality by zone, such as... Figure 1 As shown, the method includes:

[0069] SA1: Acquire sound information of the current area inside the vehicle. In this embodiment, an error sensor is set to collect sound information inside the vehicle. In order to more accurately achieve regional sound quality evaluation, error sensors are arranged in different areas to collect sound information inside the vehicle.

[0070] In one embodiment of this application, four error sensors are arranged, one each at the right ear positions of the driver, front passenger, left rear passenger, and right rear passenger. These four error sensors can collect sound information from four different areas: the driver's area, the front passenger area, the left rear passenger area, and the right rear passenger area.

[0071] SA2: Input the sound information into the sound quality evaluation model corresponding to the sound quality mode of the current area for evaluation, and obtain the sound quality score of the current area. The sound quality evaluation model is a hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm.

[0072] Based on the above technical means, a hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm was used to construct a sound quality evaluation model corresponding to different sound quality modes. The sound information of different areas in the vehicle is evaluated according to the sound quality mode requirements of users in different areas, and the evaluation accuracy is high.

[0073] In the embodiments of this application, such as Figure 2 As shown, the sound quality evaluation model is trained using the following method:

[0074] 1) Collect noise samples.

[0075] Due to the diverse needs of customers for in-vehicle sound quality, sound quality evaluation models under different sound quality modes can be built for various sound quality dimensions such as comfort, dynamic feel, and sportiness.

[0076] In this embodiment, noise signal acquisition experiments are conducted on passenger vehicles to collect noise samples under different operating conditions, on different types of road surfaces, at different test locations inside the vehicle, at different speeds, and on different vehicles. In this embodiment, different operating conditions include constant speed and acceleration; different types of road surfaces include cement roads, rough asphalt roads, and smooth asphalt roads; different test locations inside the vehicle are set according to the zoning of the vehicle interior during the sound quality evaluation process; different vehicles include vehicles from different manufacturers and brands, with prices within a preset range. The noise samples collected through the above technical means can reflect the diverse noise generated by the vehicle during actual use, and the different test locations inside the vehicle can reflect the noise situation in different areas of the vehicle interior. Such noise samples are used for model training, resulting in a more accurate model.

[0077] In one embodiment of this application, ten passenger vehicles priced between 150,000 and 300,000 yuan were selected as test vehicles. Taking noise sample collection under constant speed conditions as an example, the test was conducted on an experimental performance track. Different test positions inside the vehicle were set as the driver's position and the rear passenger seat. The test road surface included three types of road surfaces: cement road, rough asphalt road, and smooth asphalt road. The test condition was constant speed, with speed intervals of 10-20 km / h between different speeds. In this embodiment, the operating speeds were 20 km / h, 40 km / h, 60 km / h, 80 km / h, 100 km / h, and 120 km / h. Noise samples were collected during the test using a microphone or an artificial head. If a microphone was used, it could be positioned at the right ear of the driver and the right ear of the rear passenger. If an artificial head is used, it is placed in the driver's seat and the rear seat of the front passenger seat. The artificial head is positioned to simulate a normal passenger sitting posture, fixed and leaning against the seat with the tip of the artificial head aligned with the center line of the seat. The vertical height from the center of the artificial head's ears to the seat is 700mm, with a permissible height deviation of 20mm. Each test data acquisition time is 30 seconds, the sampling frequency is 25600Hz, and the frequency resolution is 2Hz. Each operating condition is tested at least three times. After data acquisition, a validity check is performed, and the set of data with good consistency is selected as the sample data. The above test includes 2 measuring points, 3 road surface types, and 6 operating conditions. Each vehicle can obtain 36 test samples, and a total of 360 noise samples under constant speed conditions are obtained from 10 vehicles. In the actual sampling process, the duration of each noise sample can be set according to requirements.

[0078] 2) Classify the noise samples according to the sound quality dimensions corresponding to different sound quality modes to obtain the noise samples corresponding to each sound quality mode. In the embodiments of this application, the sound samples under steady-state conditions such as uniform speed are generally used as the evaluation samples for the comfort sound quality dimension, and the sound samples under non-steady-state conditions such as acceleration are used as the evaluation samples for the dynamic and motion sound quality dimensions.

[0079] 3) Scoring and labeling the noise samples corresponding to each sound quality mode. After obtaining the noise samples, the evaluators score the noise samples in the sound quality evaluation room using a sound playback system consisting of headphones, a sound card, and a computer. This invention uses a 10-level scoring method to score the in-vehicle sound quality, as shown in Table 1. A sound quality score greater than 7 points represents satisfactory, and a score greater than 8 points represents good.

[0080]

[0081] Table 1

[0082] 4) Divide the noise samples and their corresponding scores for each sound quality mode into training and validation sets. Generally, the noise samples are divided into training and validation sets in an 8:2 ratio. Since the number of samples obtained through experiments is relatively small, to increase the input sample size, for noise samples under steady-state conditions, each noise sample can be truncated to a signal of a preset time length. For noise samples under non-steady-state conditions, since the noise samples themselves are unstable, the sample size can be increased by collecting more samples, or the collected samples can be truncated to longer durations, and each sample can be scored and labeled separately. For more accurate scoring, samples under non-steady-state conditions are generally truncated to noise samples longer than 4 seconds, as longer than 4 seconds better conforms to human ear evaluation standards. In the steady-state noise sample of this application, each of the aforementioned noise samples can be truncated into a 1s signal. Thus, each of the aforementioned 30s noise samples can be truncated into 30 sound samples with a duration of 1s. These 30 samples use the same scoring value, and a total of 10,800 noise samples can be obtained.

[0083] 5) The hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm is trained using training set and validation set to obtain the sound quality evaluation model corresponding to each sound quality mode.

[0084] In this embodiment, the hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and a regression output layer. Figure 3As shown, after the input layer are convolutional layers and pooling layers, which are stacked alternately. A fully connected layer maps the features learned by the convolutional and pooling layers to the sample label space. The last layer is the SVR regression output layer. The prediction model constructed using these techniques possesses both the adaptive feature extraction capabilities of convolutional neural network algorithms and the advantages of support vector regression algorithms in small data sample scenarios. It is more suitable for scenarios with high noise sample acquisition difficulty, such as in-vehicle sound quality evaluation.

[0085] The model is trained by taking the time-frequency maps of noise samples from the training and validation sets as input and the corresponding score labels of the noise samples as output. In this embodiment, the hybrid prediction model includes four convolutional layers and pooling layers, wherein: the number of convolutional kernels in the convolutional layers are 12, 24, 36, and 48, the kernel size is 3×3, the stride is 1, and the activation function is ReLU; the pooling layers use the max pooling method, the kernel size is 2×2, and the stride is 2. The parameters of the hybrid prediction model are set as follows: the initial learning rate is set to 0.0001, the minimum batch size is set to 8, the total number of iterations is set to 400 epochs, and the parameters of the hybrid prediction model are optimized using the Stochastic Gradient Descent with Momentum (SGDM) optimization algorithm.

[0086] Based on the above technical means, sound quality evaluation models corresponding to different sound qualities can be trained and used to achieve sound quality evaluation under different sound quality modes.

[0087] In practical applications, the trained model needs to have its prediction accuracy evaluated. In this embodiment, the root mean square error (RMSE) and mean absolute error (MAE) are used as the evaluation criteria for the model's prediction accuracy.

[0088] Generally speaking, the smaller the root mean square error (RMSE) and mean absolute error (MSE), the higher the model's prediction accuracy. The formulas for calculating RMSE and MAE are:

[0089]

[0090]

[0091] In the formula: y i It is a subjective evaluation score for validating a set of noisy samples. It is the prediction score of the hybrid prediction model for noisy samples in the validation set, where N is the total number of samples.

[0092] If the calculated root mean square error (RMSE) and mean absolute error (MSE) do not meet the requirements, the parameters of the hybrid prediction model need to be adjusted and the model retrained until the requirements are met.

[0093] In the above embodiment, the sound quality evaluation model corresponding to the steady-state operating condition, which was trained using the training set of 10,800 noise samples, showed that the root mean square error and mean absolute error of the prediction results on the validation set were both less than 0.1, indicating that the established CNN-SVR hybrid prediction model had high accuracy and could accurately predict the sound quality inside the vehicle.

[0094] In practice, data can be read in MATLAB or Python, and the model can be trained according to the parameters set. If the accuracy does not meet the requirements, the parameters need to be adjusted and optimized, and training can continue until the model achieves the required accuracy.

[0095] A second aspect of the present invention provides an in-vehicle sound quality zoning evaluation system, such as... Figure 4 As shown, the system includes:

[0096] The data acquisition unit is used to acquire sound information of the current area inside the vehicle. In this embodiment, the data acquisition unit directly acquires the sound information collected and transmitted by the error sensor, or the sound information collected by the error sensor is stored in a designated area, and the data acquisition unit reads the sound information from the designated area.

[0097] The sound quality scoring unit is used to input sound information into the sound quality evaluation model corresponding to the sound quality mode of the current area for evaluation, and obtain the sound quality score of the current area.

[0098] The above-mentioned technical means can be used to evaluate the sound information collected in the current area of ​​the vehicle. The evaluation is based on a hybrid prediction model of deep convolutional neural network algorithm and support vector regression algorithm. Different sound quality modes correspond to different sound quality evaluation models, and the evaluation accuracy is high.

[0099] A third aspect of the present invention provides a method for active control of in-vehicle sound quality zoning, such as... Figure 5 As shown, the method includes:

[0100] B1: Obtain the sound quality score for the current area according to the in-vehicle sound quality zoning evaluation method described above. In this embodiment, the active sound quality control is based on the sound quality mode set by the user. If the user does not set a specific mode, the initial sound quality mode is generally the universal comfort mode. After activating the active in-vehicle sound quality control, in-vehicle sound information is collected and evaluated using the aforementioned method to obtain the sound quality score for the current area.

[0101] B2: Compares the current area's sound quality score with the target score corresponding to the current area's sound quality mode. Different sound quality modes have different target scores. Users can also set the target score according to their needs, such as setting it to 7 or 8 points.

[0102] B3: If the sound quality score is lower than the target score, the active noise control system will actively control the noise in the current area; if the sound quality score is not lower than the target score, no active noise control will be performed.

[0103] B4: Repeat steps B1-B3 above until the sound quality score of the current area is not lower than the target score.

[0104] The above technical solution can achieve sound quality scoring in different areas of the vehicle. Then, based on the sound quality scores and the target score of the set sound quality mode, active noise control can be performed in the current area. This efficiently achieves the evaluation and improvement of in-vehicle sound quality, allowing different users to have different private acoustic environments according to their own needs within the same space, thus satisfying the personalized driving experience of different drivers and passengers.

[0105] In this embodiment of the application, controlling the active noise control system to actively control noise in the current area includes:

[0106] Acquire a reference signal and sound information of the current area. The reference signal is the noise source signal; in this embodiment, the vibration signal of the engine block, chassis, etc., is used as the reference signal.

[0107] The secondary sound source control signal is calculated based on the reference signal and the sound information of the current area.

[0108] The secondary sound source is controlled to generate secondary sound waves according to the secondary sound source control signal.

[0109] Based on the aforementioned technical means, the secondary sound source is controlled to emit sound according to the reference signal collected from the vehicle and the sound information of the current area. The sound waves emitted by the secondary sound source interfere with and cancel out the noise in the current area, thereby improving the sound quality score of the current area.

[0110] In this embodiment of the application, calculating the secondary sound source control signal based on the reference signal and the sound information of the current area includes:

[0111] The secondary sound source control signal is calculated using an artificial neural network algorithm based on the reference signal and the sound information of the current area.

[0112] Using the above-mentioned technical means, an artificial neural network algorithm is used to calculate the control signal of the secondary sound source. The artificial neural network algorithm has strong adaptive learning ability and nonlinear characteristics, which can better calculate the weights of the control algorithm and improve the sound quality inside the vehicle.

[0113] Active noise control, such as Figure 6 As shown. At the beginning of control, the sound information of the current area is both the primary noise signal d(n) and the in-vehicle error signal. The artificial neural network algorithm calculates the secondary sound source control signal processed by the controller based on the reference signal x(n) and the in-vehicle error signal e(n). This secondary sound source control signal controls the secondary sound source to emit sound. The generated secondary sound wave is transmitted through the secondary channel and interferes with and cancels out the original noise transmitted through the primary channel in the vehicle, forming the canceled in-vehicle sound information, which is the in-vehicle error signal e(n).

[0114] A fourth aspect of the present invention provides an in-vehicle sound quality zoning active control device, the device comprising:

[0115] The in-vehicle sound quality zoning evaluation system is used to obtain the sound quality score of the current area;

[0116] The judgment unit is used to compare the sound quality score of the current area with the target score corresponding to the sound quality pattern of the current area;

[0117] The active control unit is used to control the active noise control system to actively control noise in the current area when the sound quality score is lower than the target score.

[0118] According to the above technical solution, the device can achieve sound quality scoring in different areas of the vehicle, and then, based on the sound quality scores and the target score of the set sound quality mode as the standard, actively control the noise in the current area, efficiently realize the evaluation and improvement of the sound quality in the vehicle, so that different users can have different private acoustic environments according to their own needs in the same space, and meet the personalized driving experience of different drivers and passengers.

[0119] In this embodiment of the application, the active noise control system includes:

[0120] Multiple vibration sensors are placed at the locations where vehicle noise is generated to collect reference signals;

[0121] Multiple secondary sound sources are arranged in different areas of the vehicle to generate secondary sound waves according to the control signals of the secondary sound sources;

[0122] Multiple error sensors are deployed in different areas of the vehicle to collect sound information from different areas inside the vehicle;

[0123] The controller is used to generate control signals for the secondary sound source.

[0124] Through the above-mentioned technical means, accurate reference signals and corresponding regional sound information can be collected, enabling targeted regional sound quality control.

[0125] Active noise control systems, also known as active noise control systems, refer to systems that generate continuous sound waves in a spatial sound field to interfere with noise within the sound field, thereby achieving active noise reduction or active sound generation. Active noise control systems can be classified into three types based on the different input signals to the controller: feedforward, feedback, and hybrid control. Feedforward systems, due to the presence of a reference signal, offer good control performance and have moderate structural complexity; therefore, this invention selects a feedforward system as the active noise control system. The key to a feedforward system is acquiring a suitable reference signal. Since engine noise and tire / road noise are the main sources of noise inside the vehicle, this invention places vibration sensors in locations such as the engine block and chassis as reference signals. Based on the number of secondary sound sources and error sensors, active noise control systems can be divided into single-channel and multi-channel systems. Multi-channel systems offer good noise control performance; therefore, this invention selects a multi-channel system. Error sensors can be placed in locations such as the driver's or passenger's left or right ear inside the vehicle, while secondary sound sources can be placed in locations such as the doors, ceiling, and headrests. The placement of secondary sound sources and error sensors should be selected based on the actual project conditions, comprehensively considering factors such as effectiveness and cost.

[0126] This invention places one error sensor at each of the following locations: the driver's right ear position, the front passenger's right ear position, the left rear passenger's right ear position, and the right rear passenger's right ear position, for a total of four error sensors. Additionally, it places one secondary sound source at each of the four doors (left front door, right front door, left rear door, right rear door), and at each of the following headrest positions: the driver's headrest, the front passenger's headrest, the left rear passenger's headrest, and the right rear passenger's headrest, for a total of eight secondary sound sources. This is to meet the active control requirements for sound quality in different areas of the vehicle interior.

[0127] An active noise control system can adjust controller parameters based on the collected sound information, thereby changing the controller's output.

[0128] To further enhance the in-vehicle sound quality experience, in the embodiments of this application, such as Figure 7 As shown, the interior space of the vehicle is divided into four areas: the front driver's area (2), the front passenger area (3), the rear left passenger area (4), and the rear right passenger area (5). This allows for active control of the in-vehicle sound quality in each area, enabling different users to have different private acoustic environments according to their needs within the same vehicle. For example, the driver can choose the sporty sound quality mode to experience a sporty driving experience, while passengers can choose the comfort sound quality mode for a better rest, thus satisfying the personalized sound quality experience of different drivers and passengers.

[0129] In other embodiments, a sound quality control zone can be set according to the vehicle's seating arrangement or privacy settings, and the number and arrangement of vibration sensors, secondary sound sources, and error sensors can be arranged according to the set sound quality control zone.

[0130] The following describes the in-vehicle sound quality zone active control method of this application in conjunction with practical application. During use, it is assumed that the current area is the driver's area, the mode is set to comfort mode, and the driver sets a target score of 8 points. Figure 8 As shown, firstly, the sound information of the vehicle to be evaluated is collected. Then, the sound quality evaluation model corresponding to the sound quality mode of the current area is input for sound quality evaluation, resulting in a sound quality score. Assuming the obtained sound quality score is 7 points, the sound quality score is compared with the target score. Since 7 points is lower than the target score of 8 points, it is determined that the control target has not been achieved, and the active noise control system is activated to control the sound quality. Assuming the obtained sound quality score is 9 points, the control target is achieved, and this control operation ends.

[0131] The fifth aspect of the present invention provides a vehicle in which the in-vehicle sound quality zoning active control method described above is applied to perform in-vehicle sound quality zoning control.

[0132] Through the aforementioned technical means, vehicles can provide personalized driving and riding experiences for different drivers and passengers.

[0133] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the in-vehicle sound quality zoning evaluation method or the in-vehicle sound quality zoning active control method.

[0134] This invention utilizes convolutional neural networks (CNNs), which possess strong nonlinear mapping capabilities and can automatically extract image features, for in-vehicle sound quality evaluation and predictive modeling. Combining the adaptive feature extraction capabilities of CNNs with the advantages of support vector regression (SVR) in small data sample scenarios, an objective in-vehicle sound quality evaluation model (i.e., a CNN-SVR hybrid model) combining deep CNNs and SVR algorithms is established, achieving high model accuracy. Based on this, this invention evaluates the in-vehicle sound quality using the hybrid model. If the score does not reach the target score, an active in-vehicle sound quality control system is implemented to bring it up to the target score, efficiently achieving the evaluation and improvement of in-vehicle sound quality. To further enhance the in-vehicle sound quality experience, this invention also introduces a regional active sound quality control function, allowing different users to have different private acoustic environments according to their needs within the same vehicle space, satisfying the personalized sound quality experience of different drivers and passengers.

[0135] The above embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention.

Claims

1. A method for evaluating in-vehicle sound quality by zone, characterized in that, The method includes: Obtain sound information for the current area inside the vehicle; The sound information is input into the sound quality evaluation model corresponding to the sound quality mode of the current area for evaluation, and the sound quality score of the current area is obtained. The sound quality evaluation model is a hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm. The sound quality evaluation model is trained using the following method: Collect noise samples; Noise samples are classified according to the sound quality dimensions corresponding to different sound quality modes to obtain noise samples corresponding to each sound quality mode. Noise samples under steady-state conditions are used as evaluation samples for the comfort sound quality dimension, and noise samples under non-steady-state conditions are used as evaluation samples for the dynamic and motion sound quality dimensions. Scoring and labeling the noise samples corresponding to each sound quality mode; The noise samples and their corresponding scores for each sound quality mode are divided into training and validation sets. Each noise sample is truncated to a signal of a preset time length, thereby increasing the input sample size. After truncation, the time length of the noise sample under steady-state conditions is shorter than that under non-steady-state conditions. The hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm was trained using training set and validation set to obtain the sound quality evaluation model corresponding to each sound quality mode.

2. The in-vehicle sound quality zoning evaluation method according to claim 1, characterized in that, The noise samples include noise samples collected under different working conditions, on different types of road surfaces, at different test locations inside the vehicle, at different speeds, and on different vehicles.

3. The in-vehicle sound quality zoning evaluation method according to claim 1, characterized in that, The hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm includes an input layer, convolutional layer, pooling layer, fully connected layer and regression output layer.

4. A vehicle interior sound quality zoning evaluation system, characterized in that, The system includes: The data acquisition unit is used to acquire sound information of the current area inside the vehicle; The sound quality scoring unit is used to input sound information into the sound quality evaluation model corresponding to the sound quality mode of the current area for evaluation, and obtain the sound quality score of the current area. The sound quality evaluation model is trained using the following method: Collect noise samples; Noise samples are classified according to the sound quality dimensions corresponding to different sound quality modes to obtain noise samples corresponding to each sound quality mode. Noise samples under steady-state conditions are used as evaluation samples for the comfort sound quality dimension, and noise samples under non-steady-state conditions are used as evaluation samples for the dynamic and motion sound quality dimensions. Scoring and labeling the noise samples corresponding to each sound quality mode; The noise samples and their corresponding scores for each sound quality mode are divided into training and validation sets. Each noise sample is truncated to a signal of a preset time length, thereby increasing the input sample size. After truncation, the time length of the noise sample under steady-state conditions is shorter than that under non-steady-state conditions. The hybrid prediction model based on deep convolutional neural network algorithm and support vector regression algorithm was trained using training set and validation set to obtain the sound quality evaluation model corresponding to each sound quality mode.

5. A method for active zoning control of in-vehicle sound quality, characterized in that, The method includes: The in-vehicle sound quality zoning evaluation method according to any one of claims 1-3 obtains the sound quality score of the current area; Compare the sound quality score of the current area with the target score corresponding to the sound quality pattern of the current area; If the sound quality score is lower than the target score, the active noise control system will actively control the noise in the current area. Repeat the above steps until the sound quality score of the current area is not lower than the target score.

6. The in-vehicle sound quality zoning active control method according to claim 5, characterized in that, The active noise control system actively controls noise in the current area, including: Acquire reference signals and sound information for the current area; Calculate the secondary sound source control signal based on the reference signal and the sound information of the current area; The secondary sound source is controlled to generate secondary sound waves according to the secondary sound source control signal.

7. The in-vehicle sound quality zoning active control method according to claim 6, characterized in that, The secondary sound source control signal is calculated based on the reference signal and the sound information of the current area, including: The secondary sound source control signal is calculated using an artificial neural network algorithm based on the reference signal and the sound information of the current area.

8. An in-vehicle sound quality zone active control device, characterized in that, The device includes: The in-vehicle sound quality zoning evaluation system of claim 4 is used to obtain the sound quality score of the current area; The judgment unit is used to compare the sound quality score of the current area with the target score corresponding to the sound quality pattern of the current area; The active control unit is used to control the active noise control system to actively control noise in the current area when the sound quality score is lower than the target score.

9. The in-vehicle sound quality zone active control device according to claim 8, characterized in that, The active noise control system includes: Multiple vibration sensors are placed at the locations where vehicle noise is generated to collect reference signals; Multiple secondary sound sources are arranged in different areas of the vehicle to generate secondary sound waves according to the control signals of the secondary sound sources; Multiple error sensors are deployed in different areas of the vehicle to collect sound information from different areas inside the vehicle; The controller is used to generate control signals for the secondary sound source.

10. A vehicle, characterized in that, The vehicle uses the in-vehicle sound quality zoning active control method as described in any one of claims 5-7 to perform in-vehicle sound quality zoning control.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the in-vehicle sound quality zoning evaluation method as described in any one of claims 1-3 or the in-vehicle sound quality zoning active control method as described in any one of claims 5-7.

Images