Method, system and electronic device for analyzing in-vehicle noise

Abstract

The application provides an in-vehicle noise analysis method and system and electronic equipment, comprising: obtaining an analysis request for analyzing in-vehicle noise; wherein the analysis request comprises multiple sets of noise data in a target working condition, vehicle parameters in the corresponding target working condition, and a target data analysis method; processing the multiple sets of noise data and the corresponding vehicle parameters to obtain a set of target noise data; based on the target data analysis method and the vehicle parameters, analyzing the target noise data to obtain an analysis result of the in-vehicle noise; wherein the target data analysis method comprises one or more of the following: an AI human ear weighting analysis method, a sound quality analysis method, an order analysis method, a frequency domain masking analysis method, and an AI-CNN neural network analysis method. The application improves the accuracy of the sound quality score of the in-vehicle noise and the accuracy of the motor whistling problem positioning.

Description

Technical Field

[0001] This invention relates to the field of automotive noise testing technology, and in particular to a method, system, and electronic device for analyzing in-vehicle noise. Background Technology

[0002] With the rapid iteration of intelligence and performance of new energy vehicles, users' requirements for driving quality are increasing day by day, and in-vehicle noise control has become a core issue in the industry. Data shows that high-frequency motor whistling has ranked among the top three user complaints for three consecutive years, directly affecting driving comfort and brand reputation. The problems caused by motor whistling are particularly prominent. Currently, the commonly used noise testing methods in the industry include: (1) NVH (Noise, Vibration, Harshness) engineers drive the vehicle to conduct subjective evaluations, and combine years of experience to obtain the sound quality score in the vehicle. (2) For the motor whistling problem, professional noise acquisition sensors are used to measure data. After completing the problem condition data collection, noise signals are extracted and analyzed to check the noise level of the main order signals.

[0003] However, different engineers have varying standards for subjective evaluation, and individual differences exist in the sensitivity of the human ear to feedback. Therefore, a unified quantitative scoring standard cannot be provided, leading to biases in subjective evaluation results. Furthermore, subjective perceptions do not perfectly correspond to objective measurement data, which can hinder the localization of feedback noise issues, resulting in situations where objective measurement data scores are low while subjective perception scores are high, or vice versa. Finally, in the analysis of objective measurement data, the low weighting of mid-to-high frequency noise on human ear sensitivity can distort objective measurement data, leading to misleading definitions of feedback problems by engineers. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method, system and electronic device for analyzing in-vehicle noise, so as to improve the accuracy of sound quality scoring of in-vehicle noise and the accuracy of locating motor whine problems.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a method for analyzing in-vehicle noise, comprising: acquiring an analysis request for analyzing in-vehicle noise; wherein the analysis request includes multiple sets of noise data in the vehicle under target operating conditions, vehicle parameters under the corresponding target operating conditions, and a target data analysis method; processing the multiple sets of noise data and the corresponding vehicle parameters to obtain a set of target noise data; and analyzing the target noise data based on the target data analysis method and the vehicle parameters to obtain the analysis result of in-vehicle noise; wherein the target data analysis method includes one or more of the following: AI human ear weighting analysis method, sound quality analysis method, order analysis method, frequency domain masking analysis method, and AI-CNN neural network analysis method.

[0006] Optionally, based on the target data analysis method and vehicle parameters, the target noise data is analyzed to obtain the analysis results of the in-vehicle noise, including: if the target data analysis method includes AI human ear weighting analysis, then based on a pre-trained AI human ear weighting model, the target noise data is analyzed to obtain the frequency and magnitude of the target noise data; if the target data analysis method includes sound quality analysis, then based on a pre-trained sound quality analysis model, the target noise data is analyzed to obtain a sound quality score for the target noise data; if the target data analysis method includes order analysis, then Fourier transform is performed on the target noise data, and based on the Fourier transform result and vehicle parameters, the order features of the noise data are extracted to obtain the order analysis results of the target noise data; if the target data analysis method includes frequency domain masking analysis, then the target noise data is calculated to obtain the frequency domain masking analysis results of the target noise data; if the target data analysis method includes AI-CNN neural network analysis, then based on a pre-trained AI-CNN neural network model, feature extraction and classification recognition are performed on the target noise data to obtain the motor whine result in the target noise data.

[0007] Optionally, based on a pre-trained sound quality analysis model, sound quality analysis is performed on the target noise data to obtain a sound quality score for the target noise data, including: performing psychoacoustic analysis on the target noise data to obtain psychoacoustic parameters of the target noise data; inputting the psychoacoustic parameters into the pre-trained sound quality analysis model to obtain a sound quality score for the target noise data; wherein, the psychoacoustic parameters include at least: the decibel value, loudness, sharpness, and roughness of the target noise data.

[0008] Optionally, the target noise data is calculated to obtain the frequency domain masking analysis results of the noise data, including: converting the target noise data into a three-dimensional array form according to a preset order and a preset resolution; calculating the energy of the target order within a specified bandwidth and the energy within a 1 / 3 octave band bandwidth based on the three-dimensional array; wherein the target order is: the number of motor pole pairs, the first order of the main motor, the first, second, and third harmonics of the first order of the transmission, and the first and second harmonics of the number of motor slots; the ratio of the energy of the target order within the specified bandwidth to the energy within a 1 / 3 octave band bandwidth is determined as the frequency domain masking analysis result of the target noise data.

[0009] Optionally, the method further includes: acquiring test audio data; wherein the test audio data includes: multiple sets of single-frequency tones of different magnitudes and frequencies and multiple sets of combined tones composed of single-frequency tones; using each set of single-frequency tones and combined tones in the test audio data as a sound source, and collecting the reverberation noise and artificial head noise of each sound source in a preset experimental environment; using the frequency and magnitude of the reverberation noise of different sound sources as different input data, and using the frequency and magnitude of the corresponding artificial head noise as output labels, iteratively training the first neural network model to obtain an AI human ear weighting model.

[0010] Optionally, the method further includes: obtaining subjective scores for each group of single-frequency tones and combined tones in the test audio data; performing psychoacoustic analysis on the test audio data to obtain psychoacoustic parameters of the test audio data; using the psychoacoustic parameters as input data and the subjective scores as output labels to iteratively train the second neural network model to obtain a sound quality analysis model.

[0011] Optionally, vehicle parameters include: the vehicle's motor speed signal; processing multiple sets of noise data and corresponding vehicle parameters to obtain a set of target noise data, including: for each set of noise data, calculating the A-weighted total sound pressure level of the noise data to obtain the sound pressure level sequence corresponding to the noise data; dividing the motor speed signal into multiple speed intervals based on a preset speed range; mapping each A-weighted total sound pressure level in the sound pressure level sequence to the corresponding speed interval, and calculating the average value of the A-weighted total sound pressure level in each speed interval; obtaining a speed-sound pressure level curve based on each speed interval and the average value of the corresponding A-weighted total sound pressure level; calculating the consistency of each set of noise data based on the speed-sound pressure level curve of each set of noise data, and determining the set of noise data corresponding to the speed-sound pressure level curve with the highest consistency as the target noise data.

[0012] Secondly, the present invention provides a vehicle interior noise analysis system, comprising: an analysis request acquisition module for acquiring an analysis request for analyzing vehicle interior noise; wherein the analysis request includes multiple sets of noise data inside the vehicle under target operating conditions, vehicle parameters under the corresponding target operating conditions, and a target data analysis method; a noise data processing module for processing the multiple sets of noise data and the corresponding vehicle parameters to obtain a set of target noise data; and a noise data analysis module for analyzing the target noise data based on the target data analysis method and the vehicle parameters to obtain the analysis result of the vehicle interior noise; wherein the target data analysis method includes one or more of the following: AI human ear weighting analysis method, sound quality analysis method, order analysis method, frequency domain masking analysis method, and AI-CNN neural network analysis method.

[0013] Thirdly, the present invention provides an electronic device including a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the steps of the method provided in any of the first aspects above.

[0014] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, performs the steps of the method provided in any of the first aspects above.

[0015] This invention brings the following beneficial effects: The in-vehicle noise analysis method, system, and electronic device provided by this invention first obtain an analysis request for analyzing in-vehicle noise; wherein, the analysis request includes multiple sets of noise data in the vehicle under target operating conditions, corresponding vehicle parameters under the target operating conditions, and a target data analysis method; then, the multiple sets of noise data and the corresponding vehicle parameters are processed to obtain a set of target noise data; finally, based on the target data analysis method and the vehicle parameters, the target noise data is analyzed to obtain the in-vehicle noise analysis result; wherein, the target data analysis method includes one or more of the following: AI human ear weighting analysis method, sound quality analysis method, order analysis method, frequency domain masking analysis method, and AI-CNN neural network analysis method; the sound quality analysis method uses a deep learning algorithm to score the sound quality of the target noise data. In the above method, the sound quality score of the target noise data is obtained by using a deep learning algorithm, which enhances the correlation between the subjective and objective aspects of the sound quality score and thus improves the accuracy of the sound quality score. At the same time, the target noise data is deeply analyzed by combining frequency domain masking analysis and AI-CNN neural network analysis, so that the analysis results are more in line with the subjective perception of the human ear, thereby enabling the rapid and accurate location of motor whistling problems.

[0016] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating a method for analyzing in-vehicle noise provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the first neural network model provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the second neural network model provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the frequency domain masking analysis results provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of the AI-CNN neural network model provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of a data processing software application system based on neural network development, provided in an embodiment of the present invention. Figure 7 A flowchart illustrating a vehicle parameter module provided in an embodiment of the present invention; Figure 8 A flowchart illustrating a test data import module provided in an embodiment of the present invention; Figure 9 A flowchart illustrating a data analysis module provided in an embodiment of the present invention; Figure 10 A flowchart illustrating a data display module provided in an embodiment of the present invention; Figure 11 A flowchart illustrating a data export module provided in an embodiment of the present invention; Figure 12 A schematic diagram of the structure of an in-vehicle noise analysis system provided in an embodiment of the present invention; Figure 13 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Currently, the commonly used noise testing methods in the industry include: (1) Engineers drive the vehicle to conduct subjective evaluations and, based on years of experience, obtain a sound quality score for the vehicle interior. (2) For motor whine problems, professional noise acquisition sensors are used to measure data. After completing the problem condition data collection, noise signals are extracted and analyzed to check the noise levels of the main order signals. However, the subjective evaluation score standards of different engineers cannot be completely the same, and there are individual differences in the sensitivity of the human ear to whine. Therefore, it is impossible to provide a unified quantitative standard score, which leads to deviations in the subjective evaluation results. At the same time, subjective feelings and objective measurement data do not completely correspond, which can cause confusion in the localization of whine noise problems. There are cases where the objective measurement data score is low and the subjective feeling score is high or the opposite. Finally, in the analysis of objective measurement data, the sensitivity of mid-to-high frequency noise to the human ear is not heavily weighted, which will cause distortion of objective measurement data and thus lead to engineers being misled in their definition of whine problems.

[0022] Based on this, the present invention provides a method, system, and electronic device for analyzing in-vehicle noise, which improves the accuracy of sound quality scoring of in-vehicle noise and the accuracy of locating motor whine problems.

[0023] To facilitate understanding of this embodiment, a method for analyzing in-vehicle noise disclosed in this invention will first be described in detail. This method can be executed by an electronic device, such as a smartphone, computer, or tablet computer. See also... Figure 1 The flowchart shown illustrates a method for analyzing in-vehicle noise, indicating that the method mainly includes the following steps S101 to S103: Step S101: Obtain an analysis request for analyzing in-vehicle noise; wherein, the analysis request includes multiple sets of noise data in the vehicle under the target operating condition, vehicle parameters under the corresponding target operating condition, and target data analysis method.

[0024] In one implementation, the target operating condition can be one that causes motor whine, such as: acceleration conditions, where gear meshing noise (e.g., 7.11th and 14.22nd order of the secondary gear) can cause low-to-mid frequency whine during vehicle acceleration (e.g., 20-50 km / h); and high-frequency whine when the vehicle speed increases to 50-90 km / h. High-load conditions, where excessive motor load can lead to increased motor vibration due to mechanical jamming or overload, causing whine. Vehicle parameters include at least: vehicle model, vehicle VIN code, motor model, and motor parameters. Among these, motor parameters are the most important data, including: the number of pole pairs for the front and rear motors (only one for a single motor is needed), the main motor's first order reduction, and the transmission's first order.

[0025] In this embodiment of the invention, noise data inside the vehicle under target operating conditions and corresponding vehicle parameters can be collected. For the collection of noise data and vehicle parameters, two high-precision microphones are placed in the driver's outer ear and the rear outer ear of the front passenger. After the high-precision microphones are placed, they are connected to a high-precision acquisition device. Once the device is connected, a sound calibrator is used to calibrate the high-precision microphones using a 1000Hz & 114dB setting. If the calibrated value differs significantly from the factory value (±5dB), the high-precision microphone is replaced.

[0026] After the high-precision microphone was calibrated, two DC levelers were connected to the high-precision data acquisition equipment. Each DC leveler had five buttons, corresponding to outputs of 1V, 2V, 3V, 4V, and 5V DC voltages. The first DC leveler represented sound pressure level, with subjective perceptions corresponding to: unacceptably noisy, noisy, acceptable, pleasant, and comfortable. The second DC leveler represented feedback, with subjective perceptions corresponding to: unacceptably harsh, harsh, acceptable, slightly noticeable, and imperceptible. Simultaneously, the engineers' subjective evaluation scores were collected via the DC levelers. Engineers selected different buttons based on their subjective perceptions, with different buttons corresponding to different subjective evaluation scores.

[0027] After the two DC gear shifters are debugged, the CAN bus signal of the whole vehicle is connected to the high-precision acquisition equipment to collect vehicle parameters such as motor speed, motor torque, vehicle speed, and throttle opening. If it is a four-wheel drive electric vehicle, the speed and torque signals of the front motor and the rear motor are collected at the same time.

[0028] In this embodiment of the invention, data acquisition is carried out on professional roads or in a four-wheel drive drum test chamber to minimize the impact of the surrounding environment on the data. Three or more sets of data are collected for each target working condition to avoid randomness.

[0029] Step S102: Process multiple sets of noise data and corresponding vehicle parameters to obtain a set of target noise data.

[0030] In one implementation, after data acquisition is completed, the data is checked for anomalies, and a set of data with high consistency is selected as the typical data for external output, i.e., the target noise data, according to the "level vs rpm" analysis method.

[0031] In practical implementation, vehicle parameters include the vehicle's motor speed signal. The "level vs rpm" analysis method includes the following process: First, for each set of noise data, calculate the A-weighted total sound pressure level of the noise data to obtain the sound pressure level sequence corresponding to the noise data.

[0032] Specifically, the noise data and motor speed signal are first preprocessed to ensure that they are strictly aligned in time. Data from unstable idling or shifting processes before, after, and during the test are removed, retaining only stable acceleration or deceleration processes. The motor speed and noise signals are then checked for obvious outliers. If any are found, smoothing methods such as moving averages or median filtering can be used, or outlier data points can be directly removed. Next, the original sound pressure time-domain signal (i.e., noise data, usually voltage values ​​in V or Pa) is framed (e.g., 1024 or 2048 points per frame, with 50% overlap). For each frame, the A-weighted total sound pressure level is calculated, resulting in a time-varying sound pressure level sequence Lp(t).

[0033] Then, based on the preset speed range, the motor speed signal is divided into multiple speed intervals.

[0034] Specifically, based on a preset speed range, such as from 800 RPM to 6000 RPM, the motor speed signal is divided into multiple speed ranges (RPM Bins).

[0035] Next, each A-weighted total sound pressure level in the sound pressure level sequence is mapped to the corresponding rotational speed range, and the average value of the A-weighted total sound pressure level in each rotational speed range is calculated.

[0036] Specifically, iterate through each point in time. t Check which speed range the corresponding motor speed value falls into, and place the Lp(t) value at that time point into the corresponding speed range. After completing the above operation for all data points, calculate the average value of all Lp values ​​within each speed range.

[0037] Then, based on the average value of each speed range and the corresponding A-weighted total sound pressure level, the speed-sound pressure level curve is obtained.

[0038] Specifically, for each set of noise data, the data pairs consisting of the center points of all speed ranges and the corresponding average A-weighted total sound pressure level are plotted on the same graph to obtain the speed-sound pressure level curve.

[0039] Finally, based on the speed-sound pressure level curve of each set of noise data, the consistency of each set of noise data is calculated, and the set of noise data corresponding to the speed-sound pressure level curve with the highest consistency is determined as the target noise data.

[0040] Specifically, consistency can be judged based on the overlap and smoothness of the speed-sound pressure level (SPL) curves, or quantitative analysis can be performed on the SPL curves to determine their consistency. Specifically, the Pearson correlation coefficient can be calculated for the average speed and A-weighted total SPL of each SPL curve. A lower Pearson correlation coefficient indicates more dispersed data points, a rougher curve, and poorer consistency. Therefore, the noise data corresponding to the SPL curve with the Pearson correlation coefficient closest to 1 is selected as the target noise data.

[0041] In this embodiment of the invention, the consistency of each set of noise data can also be obtained by calculating the difference index between the curves. Specifically, the average value of the A-weighted total sound pressure level in each speed range is selected to form a reference curve. The difference, mean absolute error, root mean square error, and standard deviation between each speed-sound pressure level curve and the reference curve at each speed point are calculated. The set of noise data with the smallest mean absolute error, root mean square error, or mean standard deviation is selected as the target noise data.

[0042] Step S103: Based on the target data analysis method and vehicle parameters, analyze the target noise data to obtain the analysis results of the in-vehicle noise.

[0043] In one implementation, the target data analysis method includes one or more of the following: AI human ear weighting analysis method, sound quality analysis method, order analysis method, frequency domain masking analysis method, and AI-CNN neural network analysis method.

[0044] The in-vehicle noise analysis method provided in this embodiment of the invention uses a deep learning algorithm to score the sound quality of the target noise data, enhancing the correlation between subjective and objective sound quality scoring and thus improving the accuracy of sound quality scoring. At the same time, by combining frequency domain masking analysis and AI-CNN neural network analysis, the target noise data is analyzed in depth, making the analysis results more consistent with the subjective perception of the human ear, thereby enabling the rapid and accurate location of motor whine problems.

[0045] In one implementation, for the aforementioned step S103, that is, when analyzing the target noise data based on the target data analysis method and vehicle parameters to obtain the analysis results of the in-vehicle noise, the following methods may be used, including but not limited to: (1) If the target data analysis method includes AI human ear weighting analysis, then the target noise data is analyzed based on the pre-trained AI human ear weighting model to obtain the frequency and magnitude of the target noise data.

[0046] In practical implementation, if the target data analysis method includes AI human ear weighting analysis, an AI human ear weighting model is pre-trained. Model training includes: First, test audio data is acquired; this test audio data includes multiple sets of single-frequency tones of different magnitudes and frequencies, as well as multiple sets of combined tones composed of single-frequency tones. Specifically, in this embodiment of the invention, approximately 2000 sets of single-frequency tones are obtained based on single-frequency tones (20Hz-10000Hz with 5Hz intervals) and a magnitude of 50dB. Then, the single-frequency tones of different magnitudes (20Hz-10000Hz) are randomly combined according to a 5Hz interval between each single frequency and a magnitude range of 30dB to 60dB for each frequency, resulting in 18000 sets of combined tones.

[0047] Then, each set of single-frequency tones and combined tones in the test audio data was used as a sound source, and the reverberation noise and artificial head noise of each sound source were collected in a preset experimental environment. Specifically, the above 2000 sets of single-frequency tones and 18000 sets of combined tones were used as sound sources, and professional equipment was used to simulate an artificial head device in the ear canal. The reverberation noise and artificial head noise of each sound source were collected in a reverberation test chamber.

[0048] Finally, the frequencies and magnitudes of reverberation noise from different sound sources were used as different input data, and the frequencies and magnitudes of corresponding artificial head noise were used as output labels to iteratively train the first neural network model, resulting in the AI ​​human ear weighting model. For details, see [link to documentation]. Figure 2 The schematic diagram of the first neural network model shown illustrates that the frequency and magnitude of the reverberation data of each sound source are used as the input data of the input layer of the first neural network model. The hidden layer of the first neural network model is twice the size of the input layer. The frequency and magnitude of the artificial head noise are used as the output labels. The first neural network model is iteratively trained, and after multiple iterations, the AI ​​human ear weighting model is obtained.

[0049] Based on this, in this embodiment of the invention, the target noise data can be input into the AI ​​human ear weighting model to obtain the frequency and magnitude of the target noise data.

[0050] (2) If the target data analysis method includes sound quality analysis, then the target noise data is analyzed based on the pre-trained sound quality analysis model to obtain the sound quality score of the target noise data.

[0051] In practical implementation, if the target data analysis method includes sound quality analysis, a sound quality analysis model is pre-trained. Model training includes: First, subjective scores were obtained for each group of single-frequency tones and combination tones in the test audio data. Specifically, the aforementioned 20,000 groups of audio data were subjectively scored by at least 10 professional NVH engineers, resulting in 200,000 sets of subjective scores.

[0052] Then, psychoacoustic analysis is performed on the test audio data to obtain its psychoacoustic parameters. Specifically, each set of audio data in the test audio data is analyzed to obtain its psychoacoustic parameters, including: decibel value, loudness, sharpness, and roughness.

[0053] Finally, using psychoacoustic parameters as input data and subjective ratings as output labels, the second neural network model is iteratively trained to obtain the sound quality analysis model. For details, see [link to documentation]. Figure 3 The schematic diagram of the second neural network model shown uses four dimensions of psychoacoustic parameters as input data, which are the four independent variables of the input layer of the second neural network model. The hidden layer of the second neural network model is twice the size of the input layer and includes eight neurons. The output layer is the subjective score of each set of audio data. After iterative training of the second neural network model, the sound quality analysis model is obtained.

[0054] Based on this, when performing sound quality analysis on target noise data using a pre-trained sound quality analysis model to obtain the sound quality score of the target noise data, firstly, psychoacoustic analysis is performed on the target noise data to obtain the psychoacoustic parameters of the target noise data; then, the psychoacoustic parameters are input into the pre-trained sound quality analysis model to obtain the sound quality score of the target noise data.

[0055] In practice, the target noise data is analyzed over time for decibel values, loudness, sharpness, and roughness to obtain psychoacoustic parameters. These parameters are then imported into a sound quality analysis model for calculation to obtain a simulated human ear sound quality score.

[0056] (3) If the target data analysis method includes order analysis, then perform Fourier transform on the target noise data, and extract the order features of the noise data based on the Fourier transform results and vehicle parameters to obtain the order analysis results of the target noise data.

[0057] In practical implementation, the vehicle parameters and target noise data are first retrieved. The vehicle parameters include motor parameters, such as the number of pole pairs, first-order main reducer, first-order transmission, and number of motor slots. Then, the target noise data is segmented by time (e.g., 0.25s / segment), and order analysis parameters are set, such as a frequency resolution of 4Hz, an order bandwidth of ±0.5 orders, and an overlap ratio of 77%. Next, a Fourier transform is performed on the target noise data based on the order analysis parameters. Based on the Fourier transform results and vehicle parameters, the order features of the noise data are extracted. Specifically, the energy of each target order (i.e., the number of motor pole pairs, first-order main reducer, first-order transmission, and number of motor slots) is integrated within a ±0.5 order bandwidth. The first, second, and third harmonics of the number of motor pole pairs, first-order main reducer, and first-order transmission, as well as the first and second harmonics of the number of motor slots, are calculated. Finally, an AI-based human ear weighting module is used to perform human ear weighting on the order amplitudes, resulting in order analysis results that conform to human hearing.

[0058] (4) If the target data analysis method includes frequency domain masking analysis, then the target noise data is calculated to obtain the frequency domain masking analysis result of the target noise data.

[0059] In practice, firstly, the target noise data is converted into a three-dimensional array according to the preset order and preset resolution; then, based on the three-dimensional array, the energy of the target order within the specified bandwidth and the energy within the 1 / 3 octave band order bandwidth are calculated respectively; finally, the ratio of the energy of the target order within the specified bandwidth to the energy within the 1 / 3 octave band order bandwidth is determined as the frequency domain masking analysis result of the target noise data.

[0060] Specifically, firstly, the target noise data is filtered using an AI human ear weighting model. Then, it is converted into a three-dimensional array data with a 0.02-order bandwidth and a 25 rpm resolution. Next, the ratio of the energy of the target order (request order band) within the specified bandwidth (request order band) to the energy within the 1 / 3 octave order band (octave order band) is calculated to obtain the frequency domain masking analysis result (W result). See [link to documentation]. Figure 4 As shown, the calculation results are summarized and saved to the cache. The target order can be the number of motor pole pairs, the first order of the main motor minus one, the first, second, and third harmonics of the first order of the drive motor, or the first and second harmonics of the number of motor slots. The formula for calculating the energy of the target order within a specified bandwidth is shown below:

[0061] The formula for calculating the energy within a 1 / 3 octave band bandwidth is shown below:

[0062] The formula for calculating the frequency domain masking analysis results is shown below: W result=

[0063] (5) If the target data analysis method includes AI-CNN neural network analysis, then based on the pre-trained AI-CNN neural network model, feature extraction and classification of the target noise data are performed to obtain the motor whistling result in the target noise data.

[0064] For a detailed implementation, please refer to the structure of the AI-CNN neural network model. Figure 5 As shown, firstly, the target noise data is filtered using an AI human ear weighting model. Then, the data is segmented at 0.25s / segment and processed according to a frequency resolution of 4Hz and an overlap ratio of 77% to obtain a two-dimensional array. Next, six different feature extraction methods (e.g., standard convolution, dilated convolution, depthwise separable convolution, transposed convolution, grouped convolution, and hybrid convolution) are used to perform convolution calculations on the processed noise data to obtain data that conforms to human hearing. After passing through an excitation layer and a pooling layer, the resonance band and motor whine problem of the target noise data are automatically identified, and the order, speed, and noise level of the motor whine problem are obtained.

[0065] Furthermore, the above method also includes: combining test audio data and subjective scores, inputting them into the program's built-in API interface; after receiving the data, the program iterates on the AI-CNN neural network model, modifying the built-in classification and recognition model. By reserving ports, the program's applicability is expanded, facilitating engineers to perform sound quality scoring and troubleshooting of feedback issues in different scenarios.

[0066] Regarding the aforementioned method for analyzing in-vehicle noise, this invention provides a data processing software application system based on neural networks, see [link to relevant documentation]. Figure 6 As shown, the system includes: a vehicle parameter module, a test data import module, a data analysis module, a data display module, and a data export module.

[0067] Vehicle parameter module: Engineers can quickly input important vehicle parameters into the system to prepare for subsequent data processing. After the parameters are submitted, the program checks whether the parameters are abnormal. If the data is abnormal, a new window will pop up to remind the engineer that the entered parameters are incorrect. Please carefully check and submit again until the vehicle parameter information is entered correctly.

[0068] If the data is normal, the program checks the parameters and activates the data import module.

[0069] Test data import module: Select the path to the original data file, first clean the original data, and then determine whether the cleaned data meets the data format of the program's built-in API interface. If the conditions are not met: a new window will pop up to remind the engineer of the error message, guide the engineer to check the format of the imported typical data, open the typical data to determine if the values ​​inside are abnormal, and terminate the program process.

[0070] If the conditions are met, read the cleaned raw data, load the data into the program cache, and activate the data analysis module after reading is complete.

[0071] Data Analysis Module: Select the modules for analyzing data on the program's UI page. The more modules selected, the longer the program will take to calculate. The specific modules can be selected according to the engineer's actual situation. The analysis modules include: AI human ear weighting analysis module, sound quality analysis module, order analysis module, frequency domain masking analysis module, AI-CNN neural network analysis module, and reverse correction program module.

[0072] After confirming the selection, the program will automatically determine whether the cached vehicle parameters and raw data meet the requirements of the selected analysis module: If the requirements are not met, a new window will pop up to remind the engineer of the error message. After the engineer confirms that there are no errors, they will resubmit the program until the required runtime environment is met.

[0073] If the following conditions are met: Data analysis will begin. Data analysis takes a long time, please be patient.

[0074] Data Display Module: After the data analysis is completed, the data display module will automatically load and display the analysis results. Engineers can choose to display different analysis results based on the analysis modules selected earlier.

[0075] Data export module: Engineers can export data in different formats according to their needs, making it easy to connect to the API interfaces of other software.

[0076] To facilitate understanding, each module will be described in detail below.

[0077] See Figure 7 The diagram shows a flowchart of a vehicle parameter module. Engineers input the vehicle model, VIN code, motor model, and motor parameters as required. Motor parameters include: the number of pole pairs for the front and rear motors (only one for a single motor is needed), the main motor's first-order decrement, and the drive's first-order. After verification, the parameters are entered into the corresponding text boxes. The program automatically checks the parameters for correctness and, upon confirmation, automatically activates the test data import module.

[0078] See Figure 8The diagram shows a flowchart of a test data import module. Clicking the "Import Data" button opens a new window; select the previously mentioned "Typical" data. Clicking the "Open File" button automatically opens the file. First, the program reads the file extension, which corresponds to different file types. The program calls the corresponding API interface according to the file type. Since the original data contains a lot of comments and text annotations from the test software, the program cleans up this irrelevant information, retaining only the pure data.

[0079] During the cleaning process, there may be abnormalities in the original data, causing the system to report an error. The program will pop up a new window to display the error message and the cleaning progress. Engineers can then correct the error based on the prompts and resubmit.

[0080] After data cleaning is complete, the program will automatically read the data. During the reading process, it will check for any anomalies. If an anomaly is detected, the reading process will immediately terminate, a new window will pop up displaying the error message, and the engineer will need to verify the original data before resubmitting. If no anomalies are found, the program will read all the cleaned data and load it into the cache, and then activate the data analysis module.

[0081] See Figure 9 The diagram illustrates a data analysis module that provides engineers with multiple analysis methods, including: AI human ear weighting, sound quality analysis, order analysis, frequency domain masking analysis, AI-CNN neural network analysis, and reverse correction procedures. Specific analysis processes can be found in the aforementioned embodiments and will not be repeated here.

[0082] After the engineer selects the analysis module according to their needs, they click the data analysis button. The program calculates in the order of the modules, and after each module completes its calculation, it saves the results to the cache and activates the data display module.

[0083] See Figure 10 The diagram illustrates a data display module. Before data display begins, the program locks the vehicle parameter module, test data import module, and data analysis module to free up memory and prevent accidental duplicate runs. The header file names of the analysis results are loaded into a dropdown menu, allowing engineers to select the desired results. The program also provides multi-image association display, where images with the same units are associated. Once any set of data has been loaded and displayed, the program automatically activates the data export module.

[0084] See Figure 11 The diagram shows a flowchart of a data export module. The program provides multiple saving formats: data can be saved as CSV, XLSX, XLSM, XML, and TXT; and the display page can be saved directly as PDF, PNG, and JPG. Engineers can choose to save one or more results in batches.

[0085] The method provided in this invention combines subjective human perception with objective data using deep learning algorithms to construct a new sound quality evaluation method, enhancing the correlation between subjective and objective factors. Based on this, by combining frequency domain masking algorithms and CNN convolutional neural network algorithms, objective data is analyzed in depth, and the calculation results are more consistent with subjective human perception, thereby assisting engineers in quickly locating feedback problems.

[0086] In addition to the aforementioned method for analyzing in-vehicle noise, this invention also provides a system for analyzing in-vehicle noise, see [link to relevant documentation]. Figure 12 The schematic diagram shown illustrates the structure of an in-vehicle noise analysis system, indicating that the system may include the following components: The analysis request acquisition module 1201 is used to acquire analysis requests for analyzing in-vehicle noise; wherein, the analysis request includes multiple sets of noise data inside the vehicle under the target operating condition, vehicle parameters under the corresponding target operating condition, and target data analysis method; The noise data processing module 1202 is used to process multiple sets of noise data and corresponding vehicle parameters to obtain a set of target noise data; The noise data analysis module 1203 is used to analyze target noise data based on target data analysis methods and vehicle parameters to obtain the analysis results of in-vehicle noise; wherein, the target data analysis methods include one or more of the following: AI human ear weighting analysis method, sound quality analysis method, order analysis method, frequency domain masking analysis method and AI-CNN neural network analysis method.

[0087] The in-vehicle noise analysis system provided in this embodiment of the invention uses a deep learning algorithm to score the sound quality of the target noise data, enhancing the correlation between subjective and objective sound quality scoring and thus improving the accuracy of sound quality scoring. At the same time, by combining frequency domain masking analysis and AI-CNN neural network analysis, the target noise data is analyzed in depth, making the analysis results more consistent with the subjective perception of the human ear, thereby enabling the rapid and accurate location of motor whine problems.

[0088] It should be noted that the system provided in this embodiment of the invention has the same implementation principle and technical effects as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment can be referred to the corresponding content in the aforementioned method embodiment. The specific numerical values ​​provided in the implementation of this invention are merely exemplary and are not intended to limit the scope of the invention.

[0089] This invention also provides an electronic device, specifically, the electronic device includes a processor and a storage device; the storage device stores a computer program, and the computer program, when run by the processor, executes the method described in any of the above embodiments.

[0090] Figure 13 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 100 includes: a processor 130, a memory 131, a bus 132 and a communication interface 133. The processor 130, the communication interface 133 and the memory 131 are connected through the bus 132. The processor 130 is used to execute executable modules, such as computer programs, stored in the memory 131.

[0091] The memory 131 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 133 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0092] Bus 132 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 13 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0093] The memory 131 is used to store programs. After receiving an execution instruction, the processor 130 executes the program. The method executed by the device for defining the flow process disclosed in any of the foregoing embodiments of the present invention can be applied to the processor 130 or implemented by the processor 130.

[0094] Processor 130 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 130 or by instructions in software form. Processor 130 may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 131, and processor 130 reads the information from memory 131 and, in conjunction with its hardware, completes the steps of the above method.

[0095] The computer program product of the readable storage medium provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the foregoing method embodiments. For specific implementation, please refer to the foregoing method embodiments, which will not be repeated here.

[0096] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0097] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for analyzing in-vehicle noise, characterized in that, include: Obtain an analysis request for analyzing in-vehicle noise; wherein, the analysis request includes multiple sets of noise data in the vehicle under target operating conditions, vehicle parameters under the corresponding target operating conditions, and target data analysis methods; wherein, while collecting the noise data, the engineer's subjective evaluation score is simultaneously collected through a DC gear selector; Multiple sets of noise data and corresponding vehicle parameters are processed to obtain a set of target noise data. For each set of noise data, the A-weighted total sound pressure level (SPL) is calculated to obtain a SPL sequence. Based on a preset speed range, the motor speed signal is divided into multiple speed intervals. Each A-weighted SPL in the SPL sequence is mapped to its corresponding speed interval, and the average value is calculated to obtain a speed-SPL curve. The consistency of the speed-SPL curves for each set of noise data is calculated, and the set of noise data corresponding to the speed-SPL curve with the highest consistency is determined as the target noise data. Based on the target data analysis method and the vehicle parameters, the target noise data is analyzed to obtain the analysis results of the in-vehicle noise; wherein, the target data analysis method includes one or more of the following: AI human ear weighting analysis method, sound quality analysis method, order analysis method, frequency domain masking analysis method, and AI-CNN neural network analysis method; the AI ​​human ear weighting analysis method is implemented through a pre-trained AI human ear weighting model, which is obtained by iteratively training by taking the frequency and magnitude of reverberation noise from different sound sources as input data and the frequency and magnitude of the corresponding artificial head noise as output labels.

2. The method according to claim 1, characterized in that, Based on the target data analysis method and the vehicle parameters, the target noise data is analyzed to obtain the analysis results of the in-vehicle noise, including: If the target data analysis method includes AI human ear weighting analysis, then the target noise data is analyzed based on the pre-trained AI human ear weighting model to obtain the frequency and magnitude of the target noise data; If the target data analysis method includes a sound quality analysis method, then based on a pre-trained sound quality analysis model, sound quality analysis is performed on the target noise data to obtain a sound quality score for the target noise data. If the target data analysis method includes an order analysis method, then a Fourier transform is performed on the target noise data, and the order features of the noise data are extracted based on the Fourier transform result and the vehicle parameters to obtain the order analysis result of the target noise data. If the target data analysis method includes frequency domain masking analysis, then the target noise data is calculated to obtain the frequency domain masking analysis result of the target noise data; If the target data analysis method includes AI-CNN neural network analysis, then based on the pre-trained AI-CNN neural network model, feature extraction and classification are performed on the target noise data to obtain the motor whistling result in the target noise data.

3. The method according to claim 2, characterized in that, Based on a pre-trained sound quality analysis model, sound quality analysis is performed on the target noise data to obtain a sound quality score for the target noise data, including: Psychoacoustic analysis is performed on the target noise data to obtain the psychoacoustic parameters of the target noise data; The psychoacoustic parameters are input into a pre-trained sound quality analysis model to obtain a sound quality score for the target noise data; wherein the psychoacoustic parameters include at least the decibel value, loudness, sharpness, and roughness of the target noise data.

4. The method according to claim 2, characterized in that, The target noise data is calculated to obtain the frequency domain masking analysis results of the noise data, including: The target noise data is converted into a three-dimensional array according to a preset order and a preset resolution. Based on the three-dimensional array, the energy of the target order within the specified bandwidth and the energy within the 1 / 3 octave band bandwidth are calculated respectively; wherein, the target order is: the number of motor pole pairs, the first order of the main motor, the first, second and third harmonics of the first order of the drive motor, and the first and second harmonics of the number of motor slots; The ratio of the energy of the target order within a specified bandwidth to the energy within the bandwidth of the 1 / 3 octave band order is determined as the frequency domain masking analysis result of the target noise data.

5. The method according to claim 2, characterized in that, Also includes: Acquire test audio data; wherein, the test audio data includes: multiple sets of single-frequency tones of different magnitudes and frequencies, and multiple sets of combined tones composed of the single-frequency tones; Using each group of single-frequency tones and combined tones in the test audio data as a sound source, the reverberation noise and artificial head noise of each sound source are collected in a preset experimental environment. The frequency and magnitude of reverberation noise from different sound sources are used as different input data, and the frequency and magnitude of the corresponding artificial head noise are used as output labels to iteratively train the first neural network model to obtain the AI ​​human ear weighting model.

6. The method according to claim 5, characterized in that, Also includes: Obtain subjective scores for each group of single-frequency tones and combined tones in the test audio data; Psychoacoustic analysis is performed on the test audio data to obtain the psychoacoustic parameters of the test audio data; Using the psychoacoustic parameters as input data and the subjective rating as output label, the second neural network model is iteratively trained to obtain the sound quality analysis model.

7. A vehicle interior noise analysis system, characterized in that, include: An analysis request acquisition module is used to acquire analysis requests for analyzing in-vehicle noise; wherein, the analysis request includes multiple sets of noise data in the vehicle under target operating conditions, vehicle parameters under the corresponding target operating conditions, and target data analysis methods; wherein, while acquiring the noise data, the subjective evaluation score of the engineer is simultaneously acquired through a DC gear shifter. The noise data processing module is used to process multiple sets of noise data and corresponding vehicle parameters to obtain a set of target noise data. Specifically, for each set of noise data, the A-weighted total sound pressure level (SPL) is calculated to obtain a SPL sequence. Based on a preset speed range, the motor speed signal is divided into multiple speed intervals. Each A-weighted SPL in the SPL sequence is mapped to its corresponding speed interval, and the average value is calculated to obtain a speed-sound pressure level curve. The consistency of the speed-sound pressure level curves for each set of noise data is calculated, and the set of noise data corresponding to the speed-sound pressure level curve with the highest consistency is determined as the target noise data. The noise data analysis module is used to analyze the target noise data based on the target data analysis method and the vehicle parameters to obtain the analysis results of the in-vehicle noise. The target data analysis method includes one or more of the following: AI human ear weighting analysis, sound quality analysis, order analysis, frequency domain masking analysis, and AI-CNN neural network analysis. The AI ​​human ear weighting analysis method is implemented through a pre-trained AI human ear weighting model, which is obtained by iteratively training the model using the frequency and magnitude of reverberation noise from different sound sources as input data and the frequency and magnitude of corresponding artificial head noise as output labels.

8. An electronic device, characterized in that, The method includes a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program is executed by the processor to perform the steps of the method described in any one of claims 1 to 6.

Images