A detection method for frequency shift and abnormal sound of an electric vehicle pedestrian warning device

Abstract

This invention discloses a method for detecting frequency shift and abnormal noise in pedestrian warning devices for electric vehicles. The method triggers audio by sending a simulated operating condition message and eliminates time delay errors by synchronizing with the underlying hardware clock. The acquired audio is sequentially filtered and scaled, and effective signal segments are extracted using short-time energy and zero-crossing rate. After phase-aligned splitting, the psychoacoustic features, harmonic distortion, and frequency shift response parameters of each subsequence are extracted. Finally, a feature local distance matrix is ​​constructed, and a dynamic programming algorithm is used to calculate the minimum cumulative cost path to output the sound defect type, while simultaneously verifying the linear mapping relationship between frequency shift response and vehicle speed. This invention effectively eliminates time delay and noise interference, achieving high-precision automatic classification of abnormal noise and accurate verification of frequency shift characteristics.

Description

Technical Field

[0001] This invention relates to the field of vehicle inspection, and in particular to a method for detecting frequency shift and abnormal noise in pedestrian warning devices for electric vehicles. Background Technology

[0002] In recent years, with the popularization of new energy electric vehicles, their quietness at low speeds compared to traditional fuel vehicles has become a safety hazard for pedestrians. Therefore, relevant regulations require electric vehicles to be equipped with pedestrian warning devices. These devices must emit corresponding warning sounds based on the vehicle's real-time operating conditions, and the frequency of the audio signal must change linearly with increasing vehicle speed. Furthermore, the sound must not exhibit distortion, sharp noises, or other abnormal noise defects. However, existing pedestrian warning device detection methods have many shortcomings: First, when collecting audio under simulated operating conditions, traditional methods mostly rely on software command triggering. There is an unavoidable millisecond-level delay error between the control message sent by the communication bus and the audio collected by the acoustic device, which makes it impossible to accurately align the vehicle speed command with the audio frequency shift state, seriously affecting the accuracy of judging frequency shift inaccuracy defects. Second, the test environment is often filled with low-frequency mechanical vibration and high-frequency electromagnetic interference. Existing methods lack targeted and efficient filtering and signal normalization methods, which easily misjudge environmental noise as abnormal sound. Finally, existing abnormal sound detection methods mostly use simple acoustic threshold comparison or subjective listening, which lacks in-depth quantitative analysis of the characteristics of human auditory perception. It is impossible to accurately identify and automatically classify high-frequency abnormal sounds or distortions caused by timing misalignment and high-frequency modulation interference, which is difficult to meet the current automotive industry's urgent need for high-precision and automated acoustic quality testing. Summary of the Invention

[0003] To address the aforementioned problems in the existing technology, the present invention aims to provide a method for detecting frequency shift and abnormal noise in an electric vehicle pedestrian warning device, comprising the following steps: Step S1: Send a control message simulating the operating conditions to the pedestrian warning device under test via the communication bus. The control message encapsulates a gear shifting command and a dynamic target value for vehicle speed, driving the pedestrian warning device under test to trigger the corresponding audio signal under multiple gears and linearly progressive vehicle speed commands.

[0004] Step S2: The time series of the audio signal is acquired using the acoustic acquisition unit. After bandpass filtering and amplitude scaling processing are performed on the time series, the short-time energy value and short-time zero-crossing rate value of the sequence are extracted. The intersection point where the short-time energy value crosses the preset high energy threshold and the short-time zero-crossing rate value is lower than the preset background noise threshold is taken as the starting frame, and the point where the short-time energy value falls back to the low energy threshold is taken as the ending frame, and the effective signal segment is obtained.

[0005] Step S3: Calculate the cross-correlation coefficient between the effective signal segment and the pre-stored standard template sequence within the sliding window. Based on the peak position of the cross-correlation coefficient, perform phase alignment and truncation on the effective signal segment to split it into multiple sub-signal sequences. Extract the psychoacoustic feature vector, harmonic distortion parameters, and frequency shift response parameters of each sub-signal sequence.

[0006] Step S4: Construct a local distance matrix between the psychoacoustic feature vector and the corresponding standard sample features. Calculate a minimum cumulative cost path that traverses the matrix and satisfies the monotonically continuous constraint using a dynamic programming algorithm. Compare the cost of the minimum cumulative cost path with a preset discreteness threshold. Output the sound defect type of the pedestrian warning device under test and simultaneously verify the linear slope correspondence between the frequency shift response parameter and the vehicle speed dynamic target value.

[0007] Furthermore, step S2, which involves performing bandpass filtering and amplitude scaling on the time series, includes: A Chebyshev Type I filter with a cutoff frequency covering the human ear's audible threshold is used to perform recursive difference operations on the original time series to remove low-frequency environmental vibrations and high-frequency electromagnetic glitches. Then, the maximum absolute value of the filtered sequence is obtained as a scaling reference factor, and the amplitude of all sampling points in the sequence is divided by the scaling reference factor to map to a preset dimensionless standard extreme value range.

[0008] Furthermore, the process of extracting psychoacoustic feature vectors in step S3 includes: Step S301: Apply a Hamming window to each sub-signal sequence and perform a fast Fourier transform to obtain the frequency domain power spectrum.

[0009] Step S302: The energy of the frequency domain power spectrum is weighted into a set of triangular filter banks whose center frequencies are nonlinearly arranged according to the Mel scale, and the output energy of each triangular filter is calculated and the logarithm is taken.

[0010] Step S303: Perform discrete cosine transform decorrelation on the obtained logarithmic energy set, and extract the low-order coefficients to form the eigenvector of Mel frequency cepstral coefficients.

[0011] Furthermore, the process of extracting psychoacoustic feature vectors in step S3 also includes: Loudness extraction: The sound pressure level of the frequency domain power spectrum is corrected by frequency weighting using a preset equal loudness profile curve. The weighted frequency coordinates are mapped to the critical frequency band. The total loudness feature value is obtained by performing exponential integration on the excitation energy in each critical frequency band. Masking feature extraction: Pure tone components and noise components in the spectrum are extracted as masking tones, and their enhancement of the masking threshold of adjacent frequencies is calculated. Energy components in the current spectrum with amplitudes lower than the masking threshold are set to zero, and only the effective excitation energy exceeding the masking threshold is retained to simulate the masking characteristics of human hearing.

[0012] Furthermore, the harmonic distortion parameter mentioned in step S3 is the total harmonic distortion. Its calculation method satisfies the following formula: Its operation logic is as follows: search for the global maximum amplitude point in the frequency domain power spectrum as the fundamental voltage component. Indexed by integer harmonics of the fundamental frequency, the corresponding local peak values ​​are extracted as harmonic voltage components. Each component is extracted and substituted into the formula to quantify the degree of nonlinear distortion.

[0013] Furthermore, the frequency shift response parameter mentioned in step S3 includes the normalized frequency shift change rate. It is obtained through the following formula: ,in, To be at the reference speed The main frequency is extracted by frequency domain peak search under the instruction. To simulate vehicle speed at the current target The measured main frequency extracted under the command; its operation logic is: calculate the current measured target simulated vehicle speed. Compared with the benchmark reference speed The difference is used as a scaling term in the denominator to calculate the absolute frequency offset per unit change in vehicle speed, and then the absolute offset is divided by the reference main frequency. Perform normalized output.

[0014] Furthermore, it also includes a harmonic component comparison and analysis step: configuring multiple sets of bandpass filters with upper and lower limit frequency ratios as fractional octave filter banks, aggregating and accumulating the power spectrum energy into each fractional octave band; performing differential comparisons on the energy distribution differences between the test signal and the standard signal in the same center frequency band, and marking abnormal energy points and their corresponding frequency band positions when the difference value is greater than the preset spectral fluctuation threshold.

[0015] Furthermore, step S4, which outputs the sound defect type, also includes sharpness analysis logic: The sharpness index is calculated based on the weighted integral of the high-frequency band energy in the total energy of the entire frequency band. When the sharpness index is determined to be greater than the preset psychoacoustic deviation threshold and the total harmonic distortion parameter is greater than the preset distortion tolerance threshold, an abnormal high-frequency modulation interference is determined to have occurred in the current sub-signal sequence, and the corresponding high-frequency abnormal noise or distortion classification label is output.

[0016] Furthermore, during the verification of the frequency shift response parameters: Calculate the normalized frequency shift rate matrix under multiple consecutive vehicle speed target commands; establish the deviation function between the normalized frequency shift rate and the preset linear target rate; if the deviation function value exceeds the preset slope tolerance threshold, it is determined that the vehicle speed-audio mapping logic of the current pedestrian warning device under test has a frequency shift inaccuracy defect, and the characteristics of the control message data frame when the defect is triggered are automatically recorded.

[0017] Furthermore, it also includes the underlying hardware clock synchronization and alignment steps: While the system executes step S1, it outputs a level-flipping signal as a hardware synchronization trigger pulse to the acoustic acquisition unit through the input / output port of the bus gateway. The acoustic acquisition unit uses the rising or falling edge of the detected level-flipping signal as the sampling zero point and hard maps the network timestamp of the control message sent by the communication bus to the starting index bit of the acoustic time sequence to eliminate the millisecond-level delay error caused by the software call.

[0018] Compared to existing technologies, the advantages of this invention are as follows: By introducing a hardware clock synchronization and alignment mechanism at the system's bottom layer, and using the level-flipping signal output by the gateway as a hardware trigger pulse, this invention hard maps the network timestamp of the control message sent by the communication bus to the starting index bit of the acoustic time sequence, eliminating the millisecond-level delay error caused by software calls at the source, and achieving a seamless match between the vehicle speed dynamic target and the trigger audio; at the same time, combined with the recursive differential operation of the Chebyshev Type I filter and the dual-threshold endpoint detection algorithm, it not only effectively eliminates low-frequency environmental vibrations and high-frequency electromagnetic glitches, but also accurately extracts effective signal segments, laying an extremely pure data foundation for subsequent high-precision acoustic feature extraction.

[0019] This invention innovatively constructs a multi-dimensional feature evaluation system that integrates psychoacoustic characteristics and total harmonic distortion. Relying on dynamic programming algorithms to calculate the minimum cumulative cost path between the effective signal and the standard template, it not only overcomes the comparison difficulties caused by signal phase shift and time axis stretching and compression, but also achieves accurate classification output of abnormal sound defect types. At the same time, combined with the original normalized frequency shift rate of change matrix and deviation function, it can automatically and with high fidelity verify the linear mapping logic between vehicle speed commands and audio main frequency shift changes, comprehensively and objectively ensuring the acoustic quality and regulatory compliance of pedestrian warning devices. Attached Figure Description

[0020] Figure 1 This is an exemplary flowchart of the detection method of the present invention.

[0021] Figure 2 This is an exemplary flowchart of the steps for extracting psychoacoustic feature vectors according to the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to specific embodiments.

[0023] like Figure 1 As shown in this embodiment, a method for detecting frequency shift and abnormal noise in an electric vehicle pedestrian warning device includes the following steps: Step S1: A control message simulating operating conditions is sent to the pedestrian warning device under test via a communication bus. The control message encapsulates a gear shifting command and a dynamic target speed value, driving the pedestrian warning device to trigger corresponding audio signals under multiple gears and linearly progressive speed commands. In one example, the communication bus is a CAN bus or an in-vehicle Ethernet bus, and the control message is sent periodically according to a preset operating condition sequence. For example, the dynamic target speed value is generated according to a piecewise linear function to cover the low-speed warning range of 0-30 km / h. In another example, the gear shifting command and the dynamic target speed value are scheduled using a unified time base, enabling the pedestrian warning device to synchronously output the corresponding audio signal at the moment of gear change, thereby improving test consistency. Step S2: The time series of the audio signal is acquired using the acoustic acquisition unit. After bandpass filtering and amplitude scaling processing are performed on the time series, the short-time energy value and short-time zero-crossing rate value of the sequence are extracted. The intersection point where the short-time energy value crosses a preset high energy threshold and the short-time zero-crossing rate value is lower than a preset background noise threshold is taken as the starting frame, and the point where the short-time energy value falls back to the low energy threshold is taken as the ending frame, thus obtaining the effective signal segment. In one example, the acoustic acquisition unit includes a high-sensitivity condenser microphone and a multi-channel acquisition device, with a sampling frequency of 44.1kHz or 48kHz. For example, the short-time energy and short-time zero-crossing rate are calculated based on a sliding window with a length of 128~512 sampling points. In one example, the background noise threshold is dynamically updated using an adaptive noise estimation algorithm to adapt to different test environments.

[0024] Step S3: Calculate the cross-correlation coefficient between the effective signal segment and the pre-stored standard template sequence within the sliding window. Based on the peak position of the cross-correlation coefficient, perform phase alignment and truncation on the effective signal segment, splitting it into multiple sub-signal sequences. Extract the psychoacoustic feature vector, harmonic distortion parameters, and frequency shift response parameters of each sub-signal sequence. In one example, the standard template sequence is acquired by a calibration prototype in a standard acoustic environment. In another example, the cross-correlation coefficient is calculated using a normalized cross-correlation function to eliminate amplitude influence. For example, the signal is cyclically shifted or truncated based on the peak position of the cross-correlation coefficient to achieve time axis alignment.

[0025] Step S4: Construct a local distance matrix between the psychoacoustic feature vector and the corresponding standard sample features. Calculate a minimum cumulative cost path that traverses the matrix and satisfies the monotonically continuous constraint using a dynamic programming algorithm. Compare the cost of the minimum cumulative cost path with a preset dispersion threshold to output the sound defect type of the pedestrian warning device under test. Simultaneously verify the linear slope correspondence between the frequency shift response parameter and the vehicle speed dynamic target value. In one example, the dynamic programming algorithm is a dynamic time warping algorithm; in another example, the dispersion threshold is obtained through statistical analysis of a large number of samples; in yet another example, the sound defect type is further subdivided based on the minimum cumulative cost path, including frequency shift anomalies, timbre distortion, and periodic instability.

[0026] Step S2, which involves performing bandpass filtering and amplitude scaling on the time series, includes: A Chebyshev Type I filter with a cutoff frequency covering the human hearing threshold is used to perform recursive differencing on the original time series to remove low-frequency environmental vibrations and high-frequency electromagnetic glitches. Then, the maximum absolute value of the filtered sequence is obtained as a scaling reference factor. The amplitude of all sampling points in the sequence is divided by this scaling reference factor and mapped to a preset dimensionless standard extreme value range. In one example, the order of the Chebyshev Type I filter is 4-8, and the passband ripple is 0.5 dB. The dimensionless range is [-1, 1] to achieve unified normalization processing of data from different batches.

[0027] like Figure 2 As shown, the process of extracting psychoacoustic feature vectors in step S3 includes: Step S301: Apply a Hamming window to each sub-signal sequence and perform a Fast Fourier Transform (FFT) to obtain the frequency domain power spectrum; in one example, the length of the Hamming window is the same as that of the sub-signal sequence, and the number of FFT points is an integer power of 2.

[0028] Step S302: The energy of the frequency domain power spectrum is weighted into a set of triangular filter banks whose center frequencies are nonlinearly arranged according to the Mel scale, and the output energy of each triangular filter is calculated and the logarithm is taken; in one example, the number of filter banks is 20 to 40.

[0029] Step S303: Perform discrete cosine transform decorrelation on the obtained logarithmic energy set, and extract the lower-order coefficients to form the eigenvector of Mel frequency cepstral coefficients. In one example, the first 12th to 13th order coefficients are extracted as the eigenvector.

[0030] The process of extracting psychoacoustic feature vectors in step S3 also includes: Loudness extraction: The sound pressure level of the frequency domain power spectrum is corrected by frequency weighting using a preset equal loudness profile curve. The weighted frequency coordinates are mapped to the critical frequency band. The total loudness feature value is obtained by performing exponential integration on the excitation energy in each critical frequency band. Masking feature extraction: Pure tone components and noise components in the spectrum are extracted as masking tones, and their enhancement of the masking threshold of adjacent frequencies is calculated. Energy components in the current spectrum with amplitudes lower than the masking threshold are set to zero, and only the effective excitation energy exceeding the masking threshold is retained to simulate the masking characteristics of human hearing.

[0031] The harmonic distortion parameter mentioned in step S3 is the total harmonic distortion. Its calculation method satisfies the following formula: Its operation logic is as follows: search for the global maximum amplitude point in the frequency domain power spectrum as the fundamental voltage component. Indexed by integer harmonics of the fundamental frequency, the corresponding local peak values ​​are extracted as harmonic voltage components. Each component is extracted and substituted into the formula to quantify the degree of nonlinear distortion.

[0032] In one example, the determination of the fundamental component employs a combination of spectral peak search and neighborhood parabolic interpolation to improve frequency resolution; for instance, a local maximum point is selected in the spectral amplitude curve, and a quadratic curve fitting is performed on its two adjacent frequency points to obtain a sub-frequency resolution fundamental frequency estimate.

[0033] In one example, a frequency search tolerance window is set during the extraction of harmonic components, for example, ±1% of the fundamental frequency range, to accommodate the slight frequency drift that exists in the actual signal. In one example, when the amplitude of a certain harmonic component is lower than the preset noise threshold, it is considered an invalid harmonic and is not included in the total harmonic distortion calculation, thereby improving noise immunity.

[0034] The frequency shift response parameter mentioned in step S3 includes the normalized frequency shift change rate. It is obtained through the following formula: ,in, To be at the reference speed The main frequency is extracted by frequency domain peak search under the instruction. To simulate vehicle speed at the current target The measured main frequency extracted under the command; its operation logic is: calculate the current measured target simulated vehicle speed. Compared with the benchmark reference speed The difference is used as a scaling term in the denominator to calculate the absolute frequency offset per unit change in vehicle speed, and then the absolute offset is divided by the reference main frequency. Normalized output is performed. In one example, the extraction of the main frequency adopts a method based on spectral peak detection and multi-frame cumulative averaging. For example, the stability of the main frequency identification is improved by superimposing the energy of the FFT results of consecutive frames. In another example, for multi-harmonic structure signals, the search frequency band is limited to the fundamental frequency candidate interval to avoid misidentification of harmonics as the main frequency.

[0035] It also includes a harmonic component comparison and analysis step: configuring multiple bandpass filters with upper and lower limit frequency ratios as fractional octave filter banks, aggregating and accumulating the power spectrum energy into each fractional octave band; performing differential comparisons on the energy distribution differences between the test signal and the standard signal in the same center frequency band, and marking abnormal energy points and their corresponding frequency band positions when the difference value is greater than the preset spectral fluctuation threshold.

[0036] Step S4, which outputs the sound defect type, also includes sharpness analysis logic: The sharpness index is calculated based on the weighted integral of the high-frequency band energy in the total energy of the entire frequency band. When the sharpness index is determined to be greater than the preset psychoacoustic deviation threshold and the total harmonic distortion parameter is greater than the preset distortion tolerance threshold, an abnormal high-frequency modulation interference is determined to have occurred in the current sub-signal sequence, and the corresponding high-frequency abnormal noise or distortion classification label is output.

[0037] During the verification of the frequency shift response parameters: Calculate the normalized frequency shift rate matrix under multiple consecutive vehicle speed target commands; establish the deviation function between the normalized frequency shift rate and the preset linear target rate; if the deviation function value exceeds the preset slope tolerance threshold, it is determined that the vehicle speed-audio mapping logic of the current pedestrian warning device under test has a frequency shift inaccuracy defect, and the characteristics of the control message data frame when the defect is triggered are automatically recorded.

[0038] This embodiment also includes an underlying hardware clock synchronization and alignment step: While the system executes step S1, it outputs a level-flipping signal as a hardware synchronization trigger pulse to the acoustic acquisition unit through the input / output port of the bus gateway. The acoustic acquisition unit uses the rising or falling edge of the detected level-flipping signal as the sampling zero point and hard maps the network timestamp of the control message sent by the communication bus to the starting index bit of the acoustic time sequence to eliminate the millisecond-level delay error caused by the software call.

[0039] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications 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 this application, and should all be included within the protection scope of this application.

Claims

1. A method for detecting frequency shift and abnormal noise in a pedestrian warning device for electric vehicles, characterized in that, Includes the following steps: Step S1: Send a control message simulating the operating conditions to the pedestrian warning device under test via the communication bus. The control message encapsulates a gear shifting command and a dynamic target value for vehicle speed, driving the pedestrian warning device under test to trigger the corresponding audio signal under multiple gears and linearly progressive vehicle speed commands. Step S2: The time series of the audio signal is acquired using the acoustic acquisition unit. After bandpass filtering and amplitude scaling processing are performed on the time series, the short-time energy value and short-time zero-crossing rate value of the sequence are extracted. The intersection point where the short-time energy value crosses the preset high energy threshold and the short-time zero-crossing rate value is lower than the preset background noise threshold is taken as the starting frame, and the point where the short-time energy value falls back to the low energy threshold is taken as the ending frame, and the effective signal segment is obtained. Step S3: Calculate the cross-correlation coefficient between the effective signal segment and the pre-stored standard template sequence within the sliding window. Based on the peak position of the cross-correlation coefficient, perform phase alignment and truncation on the effective signal segment to split it into multiple sub-signal sequences. Extract the psychoacoustic feature vector, harmonic distortion parameters, and frequency shift response parameters of each sub-signal sequence. Step S4: Construct a local distance matrix between the psychoacoustic feature vector and the corresponding standard sample features. Calculate a minimum cumulative cost path that traverses the matrix and satisfies the monotonically continuous constraint using a dynamic programming algorithm. Compare the cost of the minimum cumulative cost path with a preset discreteness threshold. Output the sound defect type of the pedestrian warning device under test and simultaneously verify the linear slope correspondence between the frequency shift response parameter and the vehicle speed dynamic target value.

2. The method for detecting frequency shift and abnormal noise in an electric vehicle pedestrian warning device according to claim 1, characterized in that, Step S2, which involves performing bandpass filtering and amplitude scaling on the time series, includes: A Chebyshev Type I filter with a cutoff frequency covering the human ear's audible threshold is used to perform recursive difference operations on the original time series to remove low-frequency environmental vibrations and high-frequency electromagnetic glitches. Then, the maximum absolute value of the filtered sequence is obtained as a scaling reference factor, and the amplitude of all sampling points in the sequence is divided by the scaling reference factor to map to a preset dimensionless standard extreme value range.

3. The method for detecting frequency shift and abnormal noise in an electric vehicle pedestrian warning device according to claim 1, characterized in that, The process of extracting psychoacoustic feature vectors in step S3 includes: Step S301: Apply a Hamming window to each sub-signal sequence and perform a fast Fourier transform to obtain the frequency domain power spectrum; Step S302: Weight the energy of the frequency domain power spectrum into a set of triangular filter banks whose center frequencies are nonlinearly arranged according to the Mel scale, calculate the output energy of each triangular filter and take the logarithm. Step S303: Perform discrete cosine transform decorrelation on the obtained logarithmic energy set, and extract the low-order coefficients to form the eigenvector of Mel frequency cepstral coefficients.

4. The method for detecting frequency shift and abnormal noise in an electric vehicle pedestrian warning device according to claim 1, characterized in that, The process of extracting psychoacoustic feature vectors in step S3 also includes: Loudness extraction: The sound pressure level of the frequency domain power spectrum is corrected by frequency weighting using a preset equal loudness profile curve. The weighted frequency coordinates are mapped to the critical frequency band. The total loudness feature value is obtained by performing exponential integration on the excitation energy in each critical frequency band. Masking feature extraction: Pure tone components and noise components in the spectrum are extracted as masking tones, and their enhancement of the masking threshold of adjacent frequencies is calculated. Energy components in the current spectrum with amplitudes lower than the masking threshold are set to zero, and only the effective excitation energy exceeding the masking threshold is retained to simulate the masking characteristics of human hearing.

5. The method for detecting frequency shift and abnormal noise in an electric vehicle pedestrian warning device according to claim 1, characterized in that, The harmonic distortion parameter mentioned in step S3 is the total harmonic distortion. Its operation method satisfies the following formula: Its operation logic is as follows: search for the global maximum amplitude point in the frequency domain power spectrum as the fundamental voltage component. Indexed by integer harmonics of the fundamental frequency, the corresponding local peak values ​​are extracted as harmonic voltage components. Each component is extracted and substituted into the formula to quantify the degree of nonlinear distortion.

6. The frequency shift response parameter in step S3 according to claim 1 includes the normalized frequency shift change rate. It is obtained through the following formula: ,in, To be at the reference speed The main frequency is extracted by frequency domain peak search under the instruction. To simulate vehicle speed at the current target The measured main frequency extracted under the command; its operation logic is: calculate the current measured target simulated vehicle speed. Compared with the benchmark reference speed The difference is used as a scaling term in the denominator to calculate the absolute frequency offset per unit change in vehicle speed, and then the absolute offset is divided by the reference main frequency. Perform normalized output.

7. The method for detecting frequency shift and abnormal noise in an electric vehicle pedestrian warning device according to claim 1, characterized in that, It also includes a harmonic component comparison and analysis step: configuring multiple bandpass filters with upper and lower limit frequency ratios as fractional octave filter banks, aggregating and accumulating the power spectrum energy into each fractional octave band; performing differential comparisons on the energy distribution differences between the test signal and the standard signal in the same center frequency band, and marking abnormal energy points and their corresponding frequency band positions when the difference value is greater than the preset spectral fluctuation threshold.

8. The method for detecting frequency shift and abnormal noise in an electric vehicle pedestrian warning device according to claim 1, characterized in that, Step S4, which outputs the sound defect type, also includes sharpness analysis logic: The sharpness index is calculated based on the weighted integral of the high-frequency band energy in the total energy of the entire frequency band. When the sharpness index is determined to be greater than the preset psychoacoustic deviation threshold and the total harmonic distortion parameter is greater than the preset distortion tolerance threshold, an abnormal high-frequency modulation interference is determined to have occurred in the current sub-signal sequence, and the corresponding high-frequency abnormal noise or distortion classification label is output.

9. The method for detecting frequency shift and abnormal noise in an electric vehicle pedestrian warning device according to claim 1, characterized in that, During the verification of the frequency shift response parameters: Calculate the normalized frequency shift rate matrix under multiple consecutive vehicle speed target commands; establish the deviation function between the normalized frequency shift rate and the preset linear target rate; if the deviation function value exceeds the preset slope tolerance threshold, it is determined that the vehicle speed-audio mapping logic of the current pedestrian warning device under test has a frequency shift inaccuracy defect, and the characteristics of the control message data frame when the defect is triggered are automatically recorded.

10. The method for detecting frequency shift and abnormal noise in an electric vehicle pedestrian warning device according to claim 1, characterized in that, It also includes the underlying hardware clock synchronization and alignment steps: While the system executes step S1, it outputs a level-flipping signal as a hardware synchronization trigger pulse to the acoustic acquisition unit through the input / output port of the bus gateway. The acoustic acquisition unit uses the rising or falling edge of the detected level-flipping signal as the sampling zero point and hard maps the network timestamp of the control message sent by the communication bus to the starting index bit of the acoustic time sequence to eliminate the millisecond-level delay error caused by the software call.

Images