Broadband noise adaptive active noise reduction control method

Abstract

The invention provides a broadband noise adaptive active noise reduction control method, which belongs to the technical field of noise reduction control, and comprises the following steps of: 1, acquiring and enhancing an original reference signal for target broadband noise; step 2, multi-channel coherence calibration is carried out based on the enhanced reference signal; step 3, generating a control signal based on the calibrated reference signal; 4, collecting an error signal based on an error microphone arranged in a vehicle compartment; and step 5, performing adaptive parameter updating based on the calibrated reference signal and error signal. The problems of slow algorithm convergence and poor noise reduction performance caused by inaccurate reference signal spectrum capture, easy pollution and poor multi-channel coherence in the existing method are solved, and efficient and stable control of broadband noise is realized.

Description

Technical Field

[0001] This invention belongs to the field of noise reduction control technology, specifically relating to a broadband noise adaptive active noise reduction control method. Background Technology

[0002] With the rapid development of industrial technology and the transportation sector, broadband noise pollution has become increasingly prominent. Among these, road noise generated by tire-road friction and mechanical operating noise during vehicle operation seriously affect driving comfort and environmental safety. Active noise control technology cancels noise by generating a control signal with the same amplitude but opposite phase as the original noise. Compared with passive noise reduction, it has advantages such as lower cost, no additional weight required, and better low-frequency noise reduction effect, making it the mainstream direction for broadband noise control.

[0003] The existing vehicle noise reduction methods for broadband noise mentioned in the technical solution with patent publication number "CN114743534A" are mostly based on the FxLMS (Filtered-xLeastMeanSquare) algorithm or the traditional FxNewton algorithm. The FxLMS algorithm is widely used due to its low computational cost and strong robustness, but its convergence speed is relatively slow. While the traditional FxNewton algorithm has improved convergence speed, it has extremely high requirements for the quality of the reference signal and the coherence of multiple channels. In practical applications, these methods have significant technical drawbacks: Reference signals are susceptible to interference: The spectral characteristics of tire-road noise are complex and dynamically changing. Existing acquisition devices (such as ordinary vibration sensors and microphones) are difficult to accurately capture the complete spectrum and are easily contaminated by irrelevant noises such as engine noise and wind noise, resulting in low spectral matching between the reference signal and the target noise, which directly affects the accuracy of control signal generation. Poor coherence in multi-channel systems: In multi-channel systems, factors such as differences in the propagation paths of reference signals in each channel and installation deviations of acquisition devices can easily lead to a decrease in signal coherence between channels, resulting in distortion of gradient calculation during algorithm iteration, which in turn causes slow convergence, unstable noise reduction performance, or even noise reduction failure. Insufficient adaptive adjustment capability: The filter parameter update mechanism of the existing algorithm is fixed and cannot be dynamically adjusted according to the reference signal quality and channel coherence, making it difficult to adapt to complex and ever-changing broadband noise environments.

[0004] Therefore, developing a broadband noise adaptive active noise reduction control method that can accurately capture the target noise spectrum, optimize multi-channel coherence, and combine fast convergence with stable noise reduction performance has significant engineering application value. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a broadband noise adaptive active noise reduction control method. This method solves the problems of inaccurate reference signal spectrum capture, susceptibility to contamination, and poor multi-channel coherence in existing methods, which lead to slow algorithm convergence and poor noise reduction performance. This method achieves efficient and stable control of broadband noise.

[0006] The present invention employs the following technical solution.

[0007] A broadband noise adaptive active noise reduction control method includes: Step 1: Acquire and enhance the original reference signal for the target broadband noise; Step 2: Perform multi-channel coherence calibration based on the enhanced reference signal; Step 3: Generate control signals based on the calibrated reference signals; Step 4: Collect error signals using error microphones installed in the vehicle compartment; Step 5: Perform adaptive parameter updates based on the calibrated reference signal and error signal.

[0008] Preferably, step 1 specifically includes: Step 1-1: Acquire the original reference signal of the target broadband noise using a multi-modal acquisition unit; Step 1-2: Perform spectrum sensing enhancement processing on the original reference signal of the target broadband noise.

[0009] Preferably, step 1-1 specifically includes: Vibration signals from tire-road friction are collected by vibration sensors installed on the vehicle's floor, while sound signals are simultaneously collected by directional microphones positioned inside the wheel arches. The original reference signal of the channel, the first Channel and sampling time is The original reference signal is denoted as ,in =1,2,..., , The sampling time.

[0010] Preferably, steps 1-2 specifically include: Spectral analysis is performed on the original reference signal for each channel to extract the characteristic frequency band of the target broadband noise, and then an adaptive notch filter is used to filter it. Irrelevant noise outside the characteristic frequency band of the target broadband noise is filtered out, and then a spectrum completion algorithm is used to repair the distorted part of the original reference signal of the target broadband noise within the characteristic frequency band, resulting in a sampled signal at time [missing information]. Enhanced reference signal Enhanced reference signal The calculation formula is as follows: ,in, and These are the lowest and highest characteristic frequencies of the target broadband noise, respectively. This is the algorithm function for an adaptive notch filter. For spectrum completion functions; Spectrum completion function The mathematical expression is as follows: ; in, Algorithm function for adaptive notch filter The processed first The channel's frequency is frequency domain signal, For the completed frequency domain signal, For the enhanced reference signal, This refers to the set of effective frequency points in the spectrum without missing frequencies after processing by the algorithm function of the adaptive notch filter. This represents the set of missing and distorted frequency points after processing by the algorithm function of the adaptive notch filter. It is the inverse Fourier transform function; Spectrum completion function The execution steps are as follows: S1: Perform frequency point detection for missing and distorted frequencies, i.e., if Then determine For missing frequency points; If the phase difference between adjacent frequency points Then determine The distortion frequency point; in, The set amplitude threshold coefficient, For the current frequency Local sliding window, This is the phase change threshold. For the modulo operator, To find the average value function, To use the current frequency The set of frequency domain signals centered at the point on the left and right is formed by taking the total number of adjacent signals on both sides of the point on the left and right as equal to the size of the local sliding window. , , For the arg function; S2: Perform spectral interpolation reconstruction, that is, for Each frequency point Find its left and right nearest effective frequency points. and The frequency point was calculated using linear interpolation. Preliminary completion value : ; S3: Perform noise model completion assistance, that is, establish a statistical model of the target broadband noise based on the spectral characteristics of the effective frequency points. , The calculation method is as follows: Constructing an amplitude model: Power spectral density fitting is used to fit the amplitude. The frequency amplitude within the range is fitted to a power function using the least squares method. ,in The proportionality coefficient of the power function. The coefficients of the power function are the power coefficients. for frequency points within The corresponding actual frequency; Constructing the phase model: Using the linear phase assumption, a phase function is constructed. ,in Due to signal propagation delay, This is the initial phase; thus ; S4: Perform weighted fusion completion, which balances the local continuity of interpolation reconstruction with the global consistency of model completion by introducing weight coefficients. , The calculation formula is: ; in, , The set weight decay coefficient; S5: Perform time-domain transformation, i.e., output the completed frequency-domain signal. Perform an inverse Fourier transform to obtain the enhanced reference signal. .

[0011] Preferably, step 2 specifically includes: Step 2-1: Calculate the coherence coefficient of the enhanced reference signal for each channel; Step 2-2: Perform coherence calibration.

[0012] Preferably, step 2-1 specifically includes: For any two channels and The coherence coefficient is calculated based on the cross-power spectral density and the self-power spectral density. , The calculation formula is as follows: ; in, For channel and The cross-power spectral density of the enhanced reference signal, and Channels and The self-power spectral density of the enhanced reference signal.

[0013] Preferably, step 2-2 specifically includes: Set coherence coefficient threshold ,like < The channel is then adjusted through linear phase compensation and amplitude equalization. Enhanced reference signal The calibrated reference signal is obtained. , The calculation formula is as follows: ; in, For phase compensation function, This is the amplitude equalization function.

[0014] Preferably, step 3 specifically includes: The control coefficients and channels of the adaptive filter based on the current moment calibrated reference signal The first generation is generated by linear weighting. The speaker channel in the first Timing control signal ,Will The first one fed into the car body One serves as a sound reproduction device for loudspeakers. The calculation formula is as follows: ; in, , This represents the total number of speaker channels. For the adaptive filter in the th Time, Number The reference channel for the first... The first speaker channel Control coefficients for step taps, Let be the tap order of the adaptive filter. The negative index extension operator is defined as: .

[0015] Preferably, step 4 specifically includes: By deploying multiple sampling locations inside the vehicle compartment The first error microphone collects the residual noise signal in the noise reduction area, and the second error microphone collects the residual noise signal in the noise reduction area. The error microphone at the first The residual noise signal collected at each time point is used as the error signal. ,in =1,2,..., .

[0016] Preferably, step 5 specifically includes: Step 5-1: Perform data accumulation judgment, that is, if the calibrated reference signal With error signal The cumulative data volume has reached If there are 1, proceed to step 5-2; otherwise, proceed to step 1. Step 5-2: Perform frequency domain signal conversion, that is, perform Fourier transform on the reference signal calibrated using the overlap-preserving method and the error signal to generate a frequency domain reference signal. With frequency domain error signal , and The calculation formula is as follows: ; in, It is the Fourier transform function; Step 5-3: Update the secondary channel transfer function, i.e., based on the calibrated reference signal and error signal, update the secondary channel transfer function online using a recursive least squares algorithm. Secondary channel transfer function The calculation formula is as follows: ; in, To update the step size, This represents a function that calculates the conjugate value of the parameter within the parentheses. For regularization parameters, For the first The speaker to the Each error microphone at frequency point The updated secondary channel transfer function, For the first The speaker to the Each error microphone at frequency point The secondary channel transfer function before the update; Step 5-4: Perform frequency domain gradient vector calculation, that is, combine the calibrated reference signal, error signal, and updated secondary channel transfer function to calculate the first... The first channel to the first Each speaker channel at frequency point frequency domain gradient vector , The calculation formula is as follows: ; in, The convergence factor; Step 5-5: Update the filter parameters, that is, update the adaptive filter's parameters based on the frequency domain gradient vector. The first channel to the first Each speaker channel at frequency point Frequency domain control parameters Then, through inverse Fourier transform, it is converted into an adaptive filter for the current... Time, Number The reference channel for the first... Control coefficients for each speaker channel , The calculation formula is as follows: ; in, , for OK The identity matrix of columns, For the adaptive filter The first channel to the first Each speaker channel at frequency point The updated frequency domain control parameters, For the adaptive filter The first channel to the first Each speaker channel at frequency point The frequency domain control parameters before the update.

[0017] The beneficial effects of the present invention are as follows, compared with the prior art: By enhancing the reference signal spectrum and calibrating the multi-channel coherence, the technical defects of poor reference signal quality and poor channel coordination in existing methods are effectively solved. Combined with the improved frequency domain adaptive algorithm, fast, accurate and stable noise reduction of broadband noise is achieved, which is especially suitable for complex broadband noise scenarios such as vehicle road noise. Attached Figure Description

[0018] Figure 1 This is a flowchart of a broadband noise adaptive active noise reduction control method described in this invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, other embodiments obtained by those skilled in the art without creative effort are all within the protection scope of this invention.

[0020] like Figure 1 As shown, the broadband noise adaptive active noise reduction control method of the present invention includes: Step 1: Acquire and enhance the original reference signal for the target broadband noise; In a preferred but non-limiting embodiment of the present invention, step 1 specifically includes: Step 1-1: Acquire the original reference signal of the target broadband noise using a multi-modal acquisition unit; In a preferred but non-limiting embodiment of the present invention, step 1-1 specifically includes: Vibration signals from tire-road friction are collected by vibration sensors installed on the vehicle's floor, while sound signals are simultaneously collected by directional microphones positioned inside the wheel arches. The original reference signal of the channel, the first Channel and sampling time is The original reference signal is denoted as ,in =1,2,..., , The sampling time is specified. The original reference signal for each channel is either the vibration signal collected by a vibration sensor or the sound signal collected by a directional microphone. The broadband noise adaptive active noise reduction control method operates on a PLC installed on the vehicle. Both the vibration sensor and the directional microphone are connected to the PLC, and the original reference signal is transmitted to the PLC.

[0021] Step 1-2: Perform spectrum sensing enhancement processing on the original reference signal of the target broadband noise.

[0022] In a preferred but non-limiting embodiment of the present invention, steps 1-2 specifically include: Spectral analysis is performed on the original reference signal for each channel to extract the characteristic frequency band of the target broadband noise (such as the 30Hz-400Hz band of vehicle road noise), and an adaptive notch filter is then applied. Irrelevant noise outside the characteristic frequency band of the target broadband noise is filtered out, and then a spectrum completion algorithm is used to repair the distorted part of the original reference signal of the target broadband noise within the characteristic frequency band, resulting in a sampled signal at time [missing information]. Enhanced reference signal Enhanced reference signal The calculation formula is as follows: ,in, and These are the lowest and highest characteristic frequencies of the target broadband noise, respectively. This is the algorithm function for an adaptive notch filter. This is the spectrum completion function. and These are the lowest and highest characteristic frequencies of the target broadband noise, which are the lowest and highest frequencies of the reference signal of the target broadband noise within its characteristic frequency band at all sampling times.

[0023] Furthermore, in a preferred but non-limiting embodiment of the present invention, in steps 1-2, the spectral completion function... This function targets the characteristic frequency bands of the enhanced reference signal (e.g., road noise from vehicles in the 30Hz-400Hz range). Through noise statistical modeling and signal correlation reconstruction, it fills in missing or distorted frequency bands in the spectrum caused by interference or sensor distortion, outputting a reference signal with a complete spectrum and consistent characteristics. (Spectrum completion function) The mathematical expression is as follows: ; in, Algorithm function for adaptive notch filter The processed first The channel's frequency is The frequency domain signal (specifically, the frequency domain signal is obtained by applying an algorithm function to an adaptive notch filter) After processing, Fourier transform is applied to obtain the result. For the completed frequency domain signal, For the enhanced reference signal, This refers to the set of effective frequency points in the spectrum without missing frequencies after processing by the algorithm function of the adaptive notch filter. This represents the set of missing and distorted frequency points after processing by the algorithm function of the adaptive notch filter. It is the inverse Fourier transform function; according to The values ​​are arranged in ascending order.

[0024] Spectrum completion function The execution steps are as follows: S1: Perform frequency point detection for missing and distorted signals, i.e., first analyze the frequency domain signal after notch filtering. Perform amplitude and phase stability analysis and define frequency point validity criteria: Execution amplitude criterion: If Then determine For missing frequency points; Phase criterion: If the phase difference between adjacent frequency points is... Then determine The distortion frequency point; in, The set amplitude threshold coefficient ( It can be set according to specific requirements, and its value range can be 0.3-0.5). For the current frequency The local sliding window (the window size of the local sliding window is an odd number, which can be set according to specific requirements, and its value range can be 5-11 frequency points). Phase change threshold ( It can be set according to specific requirements, and its value range can be 90°-120°. For the modulus (amplitude) operator, The averaging function calculates the arithmetic mean of all elements within the parentheses, thus obtaining the average signal amplitude within a local sliding window to avoid the influence of random fluctuations at individual frequency points on the judgment. To use the current frequency The set is formed by taking the frequency domain signals that are adjacent to the center and whose total number is equal to the size of the local sliding window. The frequency domain signals in the set are the elements of the set. If the current frequency point If the number of adjacent frequency domain signals on a certain side is less than half the size of the local sliding window minus one, then more of the current frequency points will be selected. The adjacent frequency domain signal on the other side is used to make up for the insufficient quantity on one side. , , The `arg` function represents the frequencies other than those that are missing or distorted. These frequencies are considered valid frequencies. The set of valid frequencies is formed by dividing the valid frequencies according to the above criteria. And the set of missing and distorted frequencies formed by the combination of missing and distorted frequencies. . and The frequency points are all based on its The values ​​are arranged in ascending order.

[0025] S2: Perform spectral interpolation reconstruction, that is, for Each frequency point Find its left and right nearest effective frequency points. (this frequency) The most recent effective frequency on the left) and (this frequency) The most recent effective frequency on the right (calculated using linear interpolation) is used to determine this frequency. Preliminary completion value : If for Each frequency point When searching for its left and right nearest effective frequency points, if there is a case where the nearest effective frequency point on one side does not exist, then... The average value of all valid frequency points is taken as the frequency point. The initial completion value. S2 utilizes the spectral continuity of adjacent effective frequency points to ensure the local smoothness of the completed signal.

[0026] S3: Perform noise model completion assistance, that is, to improve the accuracy of completion, establish a statistical model of the target broadband noise based on the spectral characteristics of the effective frequency points. , The calculation method is as follows: Constructing an amplitude model: A power spectral density (PSD) fitting method is used to fit the amplitude. The frequency amplitude within the range is fitted to a power function using the least squares method. ,in The proportionality coefficient of the power function. Let be the power coefficient of the power function, and be... frequency points within The corresponding actual frequency; Constructing the phase model: Using the linear phase assumption, a phase function is constructed. ,in For signal propagation delay (this signal propagation delay is achieved through the first...) The linear relationship between the phase and frequency of the effective frequency point of the channel is obtained by constructing a fitting equation. For the initial phase ( For the first channel (theoretical value of phase at time) thus ; S4: Perform weighted fusion completion, which balances the local continuity of interpolation reconstruction with the global consistency of model completion by introducing weight coefficients. This allows for a weighted fusion of the interpolation results and the model results. The calculation formula is: ; in, , The weight decay coefficient is set (the weight decay coefficient is set according to specific requirements, and its value range can be 1.5-2.5); this weight design makes missing frequencies closer to the effective frequency point more dependent on the interpolation result, and missing frequencies farther away from the effective frequency point more dependent on the model result. If or No, weighting coefficients do not exist. It is set according to specific requirements, and its value range is 0-1.

[0027] S5: Perform time-domain transformation, i.e., output the completed frequency-domain signal. Perform an inverse Fourier transform (IFFT) to obtain the enhanced reference signal. .

[0028] Spectrum completion function The technical effects are shown below: 1. Significantly improved spectral integrity For broadband signals such as tire-road noise, the spectrum completion function can effectively fill in the missing / distorted spectral regions left after notch filtering. The completed spectrum effectively covers the target characteristic frequency band (30Hz-400Hz), and the proportion of missing frequency points is reduced from 15%-25% before completion to less than 3%, improving the spectrum continuity by more than 80%.

[0029] 2. Reference signal quality optimization The spectral matching degree between the completed reference signal and the target broadband noise is improved from 75%-80% after notch filtering to over 92%, and the phase distortion is reduced by 60%. This provides a high-quality signal foundation for subsequent multi-channel coherence calibration and avoids misjudgment of inter-channel coherence due to missing spectrum.

[0030] 3. Guaranteed algorithm noise reduction performance In complex noise environments (such as mixed wind noise and engine noise interference), the spectrum completion function improves the signal-to-noise ratio of the reference signal by 5-7dB, thereby ensuring the accuracy of gradient calculation in the adaptive filtering algorithm. This improves the convergence speed of the algorithm by 30%-50% compared to the scheme without spectrum completion, and ultimately improves the noise reduction by 3-5dB. The noise reduction effect is most significant in the low-frequency missing band (30Hz-100Hz).

[0031] 4. High robustness This function, through a design that combines local interpolation with a global model, is suitable for both small-scale spectrum loss scenarios and continuous multi-frequency point loss situations. The adaptive adjustment mechanism of the weight coefficients ensures that the completion results remain stable under different noise intensities. In dynamic scenarios such as vehicle speed changes (30km / h-120km / h) and road surface type switching (asphalt road, cement road), the completion effect fluctuation is ≤0.5dB.

[0032] Step 2: Perform multi-channel coherence calibration based on the enhanced reference signal; In a preferred but non-limiting embodiment of the present invention, step 2 specifically includes: Step 2-1: Calculate the coherence coefficient of the enhanced reference signal for each channel; In a preferred but non-limiting embodiment of the present invention, step 2-1 specifically includes: For any two channels and The coherence coefficient is calculated based on the cross-power spectral density and the self-power spectral density. , The calculation formula is as follows: ; in, For channel and The cross-power spectral density of the enhanced reference signal, )and Channels and The self-power spectral density of the enhanced reference signal.

[0033] Step 2-2: Perform coherence calibration.

[0034] In a preferred but non-limiting embodiment of the present invention, step 2-2 specifically includes: Set coherence coefficient threshold ( The value is set according to specific requirements, and its range can be 0.6-0.8. < The channel is then adjusted through linear phase compensation and amplitude equalization. Enhanced reference signal The calibrated reference signal is obtained. , The calculation formula is as follows: ; in, For phase compensation function, This is the amplitude equalization function. The phase compensation function can be a function that executes the SPSC algorithm, and the amplitude equalization function can be a function that executes the modified constant modulus algorithm.

[0035] Step 3: Generate control signals based on the calibrated reference signals; In a preferred but non-limiting embodiment of the present invention, step 3 specifically includes: The control coefficients and channels of the adaptive filter based on the current moment calibrated reference signal The first generation is generated by linear weighting. The speaker channel in the first Time (in essence) Control signal (indicating the index of the sampling time) ,Will The first one fed into the car body Each is a sound reproduction device that functions as a loudspeaker (such as a headrest speaker or a door speaker; the sound reproduction device is connected to a PLC). The calculation formula is as follows: ; in, , This represents the total number of speaker channels; one speaker channel represents one speaker. For the adaptive filter in the th Time, Number The reference channel for the first... The first speaker channel Control coefficients for step taps, Let be the tap order of the adaptive filter. The negative index extension operator is defined as: The adaptive filter can be an LMS adaptive filter. In The non-negative index is used to obtain the true calibrated reference signal. The index is a negative number, which is then padded with zeros for extension.

[0036] Step 4: Collect error signals using error microphones installed in the vehicle compartment; In a preferred but non-limiting embodiment of the present invention, step 4 specifically includes: By deploying multiple sampling locations inside the vehicle compartment The first error microphone collects the residual noise signal in the noise reduction area, and the second error microphone collects the residual noise signal in the noise reduction area. The error microphone at the first The residual noise signal collected at each time point is used as the error signal. ,in =1,2,..., The error microphone is connected to the PLC, and it transmits the collected error signals to the PLC.

[0037] Step 5: Perform adaptive parameter updates based on the calibrated reference signal and error signal.

[0038] In a preferred but non-limiting embodiment of the present invention, step 5 specifically includes: Step 5-1: Perform data accumulation judgment, that is, if the calibrated reference signal With error signal The cumulative data volume has reached If there are 1, proceed to step 5-2; otherwise, proceed to step 1. Step 5-2: Perform frequency domain signal transformation, that is, use the overlap-save method to perform Fourier transform (FFT) on the calibrated reference signal and the error signal to generate the frequency domain reference signal. With frequency domain error signal , and The calculation formula is as follows: ; in, Essentially a frequency index, It is the Fourier transform function; Essentially, it is the first Each channel at frequency point The frequency domain reference signal, Essentially, it is the first Each error microphone at frequency point The frequency domain error signal.

[0039] Step 5-3: Update the secondary channel transfer function, i.e., based on the calibrated reference signal and error signal, update the secondary channel transfer function online using the recursive least squares (RLS) algorithm. ( Essentially, it is the first The speaker to the Secondary channel transfer function of each error microphone), secondary channel transfer function The calculation formula is as follows: ; in, To update step size ( The value is set according to specific requirements and can range from 0.01 to 0.1. This represents a function that calculates the conjugate value of the parameter within the parentheses. For regularization parameters ( The value is set according to specific requirements, and its range can be 1e-6-1e-4 (to avoid a denominator of zero). For the first The speaker to the Each error microphone at frequency point The updated secondary channel transfer function, For the first The speaker to the Each error microphone at frequency point The secondary channel transfer function before the update; No. The speaker to the The initial secondary channel transfer function of each error microphone can be obtained using the following method: (1) PLC to the first Each speaker outputs a known test signal. (e.g., pseudo-random noise PN, frequency sweep signal, with frequency band covering the target noise reduction range, such as 30Hz-400Hz). (2) No. Each error microphone simultaneously acquires the response signal. (The output of the test signal after passing through the secondary channel); (3) To and Perform Fast Fourier Transform on each, and obtain the results respectively. and Frequency domain signal and ; (4) Calculate the first The speaker to the The initial secondary channel transfer function of an error microphone The calculation formula is as follows: ,in , Essentially, it is the length of the Fast Fourier Transform.

[0040] Step 5-4: Perform frequency domain gradient vector calculation, that is, combine the calibrated reference signal, error signal, and updated secondary channel transfer function to calculate the first... The first channel to the first Each speaker channel at frequency point frequency domain gradient vector , The calculation formula is as follows: ; in, Convergence factor ( The value is set according to specific requirements and can range from 0 to 2. Step 5-5: Update the filter parameters, that is, update the adaptive filter's parameters based on the frequency domain gradient vector. The first channel to the first Each speaker channel at frequency point Frequency domain control parameters Then, it is converted into an adaptive filter through inverse Fourier transform (IFFT) at the current... Time, Number The reference channel for the first... Control coefficients for each speaker channel , The calculation formula is as follows: ; in, , for OK The identity matrix of columns, for OK The zero matrix of columns, Used to extract the first N time-domain sampling points. For the adaptive filter The first channel to the first Each speaker channel at frequency point The updated frequency domain control parameters, For the adaptive filter The first channel to the first Each speaker channel at frequency point The frequency domain control parameters before the update. Current number Time, Number The reference channel for the first... Control coefficients for each speaker channel It is a length of The array, the array's first... The element is .

[0041] To verify the technical effectiveness of this invention, a vehicle road noise simulation experimental platform was built. Wideband white noise of 30Hz-400Hz was used to simulate tire-road noise. Comparative experiments were conducted using the method of this invention, the traditional multi-channel FxLMS algorithm, and the traditional FxNewton algorithm. The experimental results are as follows: Reference signal quality: The spectral matching degree between the reference signal and the target noise of the method of the present invention reaches 92%, which is more than 30% higher than that of the traditional method (65%-70%), effectively filtering out interference such as wind noise and engine noise; Multi-channel coherence: After calibration, the average coherence coefficient of each channel reaches 0.85, which is 40% better than the traditional method (0.6-0.65), and the signal coordination between channels is significantly improved; Algorithm convergence speed: The convergence iterations of the method of this invention are 5000-8000 times, which is 2-3 times faster than the traditional FxLMS algorithm (20000-30000 times) and 1.5 times faster than the traditional FxNewton algorithm (12000-15000 times); Noise reduction performance: In the 30Hz-400Hz frequency band, the average noise reduction of the method of this invention reaches 18-22dB, which is 5-7dB higher than the traditional FxLMS algorithm (12-15dB) and 4-6dB higher than the traditional FxNewton algorithm (14-16dB); especially in the low frequency band (30Hz-100Hz), the noise reduction improvement is more significant, reaching 20-25dB; Stability: In a dynamic noise experiment lasting 2 hours, the noise reduction of the method of this invention fluctuated by ≤1dB, which is more stable than the traditional method (fluctuation of 2-3dB).

[0042] The beneficial effects of the present invention are as follows, compared with the prior art: By enhancing the reference signal spectrum and calibrating the multi-channel coherence, the technical defects of poor reference signal quality and poor channel coordination in existing methods are effectively solved. Combined with the improved frequency domain adaptive algorithm, fast, accurate and stable noise reduction of broadband noise is achieved, which is especially suitable for complex broadband noise scenarios such as vehicle road noise.

[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention without departing from the spirit and scope of the present invention. Any modifications or equivalent substitutions should be covered within the scope of protection of the claims of the present invention.

Claims

1. A broadband noise adaptive active noise reduction control method, characterized in that, include: Step 1: Acquire and enhance the original reference signal for the target broadband noise; Step 2: Perform multi-channel coherence calibration based on the enhanced reference signal; Step 3: Generate control signals based on the calibrated reference signals; Step 4: Collect error signals using error microphones installed in the vehicle compartment; Step 5: Perform adaptive parameter updates based on the calibrated reference signal and error signal.

2. The broadband noise adaptive active noise reduction control method according to claim 1, characterized in that, Step 1 specifically includes: Step 1-1: Acquire the original reference signal of the target broadband noise using a multi-modal acquisition unit; Step 1-2: Perform spectrum sensing enhancement processing on the original reference signal of the target broadband noise.

3. The broadband noise adaptive active noise reduction control method according to claim 2, characterized in that, Step 1-1 specifically includes: Vibration signals from tire-road friction are collected by vibration sensors installed on the vehicle's floor, while sound signals are simultaneously collected by directional microphones positioned inside the wheel arches. The original reference signal of the channel, the first Channel and sampling time is The original reference signal is denoted as ,in =1,2,..., , The sampling time.

4. The broadband noise adaptive active noise reduction control method according to claim 3, characterized in that, Steps 1-2 specifically include: Spectral analysis is performed on the original reference signal for each channel to extract the characteristic frequency band of the target broadband noise, and then an adaptive notch filter is used to filter it. Irrelevant noise outside the characteristic frequency band of the target broadband noise is filtered out, and then a spectrum completion algorithm is used to repair the distorted part of the original reference signal of the target broadband noise within the characteristic frequency band, resulting in a sampled signal at time [missing information]. Enhanced reference signal Enhanced reference signal The calculation formula is as follows: ,in, and These are the lowest and highest characteristic frequencies of the target broadband noise, respectively. This is the algorithm function for an adaptive notch filter. For spectrum completion functions; Spectrum completion function The mathematical expression is as follows: ； in, Algorithm function for adaptive notch filter The processed first The channel's frequency is frequency domain signal, For the completed frequency domain signal, For the enhanced reference signal, This refers to the set of effective frequency points in the spectrum without missing frequencies after processing by the algorithm function of the adaptive notch filter. This represents the set of missing and distorted frequency points after processing by the algorithm function of the adaptive notch filter. It is the inverse Fourier transform function; Spectrum completion function The execution steps are as follows: S1: Perform frequency point detection for missing and distorted frequencies, i.e., if Then determine For missing frequency points; If the phase difference between adjacent frequency points Then determine The distortion frequency point; in, The set amplitude threshold coefficient, For the current frequency A partial sliding window, This is the phase change threshold. For the modulo operator, To find the average function, To use the current frequency The set of frequency domain signals centered at the point on the left and right is formed by taking the total number of adjacent signals on both sides of the point on the left and right as equal to the size of the local sliding window. , , For the arg function; S2: Perform spectral interpolation reconstruction, that is, for Each frequency point Find its left and right nearest effective frequency points. and The frequency point was calculated using linear interpolation. Preliminary completion value : ； S3: Perform noise model completion, i.e., establish a statistical model of the target broadband noise based on the spectral characteristics of the effective frequency points. , The calculation method is as follows: Constructing an amplitude model: Power spectral density fitting is used to... The frequency amplitude within the range is fitted to a power function using the least squares method. ,in The proportionality coefficient of the power function. The coefficients of the power function are the power coefficients. for Frequency points within The corresponding actual frequency; Constructing the phase model: Using the linear phase assumption, a phase function is constructed. ,in Due to signal propagation delay, This is the initial phase; thus ; S4: Perform weighted fusion completion, which balances the local continuity of interpolation reconstruction with the global consistency of model completion by introducing weight coefficients. , The calculation formula is: ； in, , The set weight decay coefficient; S5: Perform time-domain transformation, i.e., output the completed frequency-domain signal. Perform an inverse Fourier transform to obtain the enhanced reference signal. .

5. The broadband noise adaptive active noise reduction control method according to claim 4, characterized in that, Step 2 specifically includes: Step 2-1: Calculate the coherence coefficient of the enhanced reference signal for each channel; Step 2-2: Perform coherence calibration.

6. The broadband noise adaptive active noise reduction control method according to claim 5, characterized in that, Step 2-1 specifically includes: For any two channels and The coherence coefficient is calculated based on the cross-power spectral density and the self-power spectral density. , The calculation formula is as follows: ； in, For channel and The cross-power spectral density of the enhanced reference signal, and Channels and The self-power spectral density of the enhanced reference signal.

7. The broadband noise adaptive active noise reduction control method according to claim 6, characterized in that, Step 2-2 specifically includes: Set coherence coefficient threshold ,like < The channel is then adjusted through linear phase compensation and amplitude equalization. Enhanced reference signal The calibrated reference signal is obtained. , The calculation formula is as follows: ； in, For phase compensation function, This is the amplitude equalization function.

8. The broadband noise adaptive active noise reduction control method according to claim 7, characterized in that, Step 3 specifically includes: The control coefficients and channels of the adaptive filter based on the current moment The calibrated reference signal The first generation is generated by linear weighting. The speaker channel in the first Timing control signal ,Will The first one fed into the car body One serves as a sound reproduction device for loudspeakers. The calculation formula is as follows: ； in, , This represents the total number of speaker channels. For the adaptive filter in the th Time, Number The reference channel for the first... The first speaker channel Control coefficients for step taps, Let be the tap order of the adaptive filter. The negative index extension operator is defined as: 。 9. The broadband noise adaptive active noise reduction control method according to claim 8, characterized in that, Step 4 specifically includes: By deploying multiple sampling locations inside the vehicle compartment The first error microphone collects the residual noise signal in the noise reduction area, and the second error microphone collects the residual noise signal in the noise reduction area. The error microphone at the first The residual noise signal collected at each time point is used as the error signal. ,in =1,2,..., .

10. The broadband noise adaptive active noise reduction control method according to claim 9, characterized in that, Step 5 specifically includes: Step 5-1: Perform data accumulation judgment, that is, if the calibrated reference signal With error signal The cumulative data volume has reached If there are 1, proceed to step 5-2; otherwise, proceed to step 1. Step 5-2: Perform frequency domain signal conversion, that is, perform Fourier transform on the reference signal calibrated using the overlap-preserving method and the error signal to generate a frequency domain reference signal. With frequency domain error signal , and The calculation formula is as follows: ； in, It is the Fourier transform function; Step 5-3: Update the secondary channel transfer function, i.e., based on the calibrated reference signal and error signal, update the secondary channel transfer function online using a recursive least squares algorithm. Secondary channel transfer function The calculation formula is as follows: ； in, To update the step size, This represents a function that calculates the conjugate value of the parameter within the parentheses. For regularization parameters, For the first The speaker to the Each error microphone at frequency point The updated secondary channel transfer function, For the first The speaker to the Each error microphone at frequency point The secondary channel transfer function before the update; Step 5-4: Perform frequency domain gradient vector calculation, that is, combine the calibrated reference signal, error signal, and updated secondary channel transfer function to calculate the first... The first channel to the first Each speaker channel at frequency point frequency domain gradient vector , The calculation formula is as follows: ； in, The convergence factor; Step 5-5: Update the filter parameters, that is, update the adaptive filter's parameters based on the frequency domain gradient vector. The first channel to the first Each speaker channel at frequency point Frequency domain control parameters Then, through inverse Fourier transform, it is converted into an adaptive filter for the current... Time, Number The reference channel for the first... Control coefficients for each speaker channel , The calculation formula is as follows: ； in, , for OK The identity matrix of columns, For the adaptive filter The first channel to the first Each speaker channel at frequency point The updated frequency domain control parameters, For the adaptive filter The first channel to the first Each speaker channel at frequency point The frequency domain control parameters before the update.

Images