A multi-channel active noise reduction system and method for improving sound quality in a vehicle

Abstract

The present invention discloses a multi-channel active noise reduction system and method for improving in-vehicle sound quality. The system comprises an audio transceiver module, an audio data transmission module, a noise signal analysis and preprocessing module, a reference signal generation module, a FELMS-based multi-channel noise reduction algorithm module, and a basic parameter setting module. The noise signal analysis and preprocessing module performs noise signal analysis and preprocessing on the noise source signal, then sends the reference signal required for control to the noise reduction algorithm module. After performing calculations based on the received error signal, the module generates an anti-noise signal to achieve noise cancellation. The present invention can analyze the sound quality characteristics of in-vehicle noise and conduct targeted control. This improves the in-vehicle sound quality while meeting the requirements of a larger interior noise reduction space, thereby enhancing the driving experience.

Description

Technical Field

[0001] The present invention relates to a multi-channel active noise reduction system and method for improving sound quality in a vehicle, which is used to achieve noise control and improve the acoustic environment in the vehicle. Background Art

[0002] Passive noise reduction, such as installing sound insulation materials, adding sound-absorbing devices, and using panel materials to suppress vibrations, is simple to operate and easy to implement. It has a significant effect on controlling high-frequency noise inside the car, but it is not effective in suppressing medium and low-frequency noise inside the car, especially low-frequency noise that has a greater impact on human hearing. The effect is even less significant in fuel-powered vehicles powered by engines.

[0003] As the automotive industry's "New Four Modernizations" trend gradually advances, traditional active noise control technologies primarily focus on reducing sound pressure levels at target noise reduction points, without further consideration of the human ear's auditory characteristics, thus failing to truly enhance auditory comfort for drivers and passengers. Furthermore, the vehicle cockpit is a three-dimensional space, and traditional single-channel active noise control technologies fail to meet the noise reduction needs of drivers and passengers for a larger area, nor do they effectively integrate with onboard systems. Summary of the Invention

[0004] The present invention aims to address the shortcomings of the above-mentioned existing technologies and proposes a multi-channel active noise reduction system and method for improving the sound quality of the vehicle, so as to analyze the sound quality characteristics and carry out targeted control of the noise inside the vehicle, thereby improving the sound quality inside the vehicle while meeting the requirements of noise reduction in a larger space in the vehicle, thereby enhancing the driving experience.

[0005] In order to solve the above technical problems, the following technical solutions are adopted:

[0006] The present invention provides an in-vehicle multi-channel active noise reduction system for improving sound quality, comprising: an audio transceiver module, an audio data transmission module, a noise signal analysis and preprocessing module, a reference signal generation module, a FELMS-based multi-channel noise reduction algorithm module, and a basic parameter setting module; wherein the audio transceiver module comprises: a reference signal receiving unit, an error signal receiving unit, and an anti-noise signal sending unit; the noise signal analysis and preprocessing module comprises: a time / frequency data conversion unit and a sound quality parameter calculation and analysis unit;

[0007] The basic parameter setting module sets the step size factor μ, the acquisition time L, and the four adaptive filter weight coefficient matrices {H cb (n)|c=1,2;b=1,2}, where H cb (n) represents the adaptive filter weight coefficient matrix at the cth row and bth column at the nth moment;

[0008] The reference signal receiving unit obtains the noise source signal x(n) at the nth moment and sends it to the time / frequency data conversion unit through the audio data transmission module;

[0009] The time / frequency data conversion unit performs spectrum analysis on the noise source signal x(n) at the nth moment using a fast Fourier transform method to obtain frequency domain amplitude spectrum and energy spectrum information of the noise source signal x(n) at the nth moment and sends the information to the sound quality parameter calculation and analysis unit;

[0010] The sound quality parameter calculation and analysis unit processes the frequency domain amplitude spectrum and energy spectrum information using the sound quality parameter calculation model to obtain the critical frequency band energy and sound pressure level of the noise source signal x(n) at the nth moment, thereby calculating the sound quality parameter eigenvalue matrix of the noise source signal x(n) at the nth moment according to the sound pressure level, and selecting the frequency values ​​and amplitudes corresponding to the two largest eigenvalues ​​from the sound quality parameter eigenvalue matrix based on the frequency domain amplitude spectrum and energy spectrum information and sending them to the reference signal generation module;

[0011] The reference signal generation module generates two reference signal matrices {X c (n)|c=1,2} and sent to the FELMS multi-channel noise reduction algorithm module, where X c (n) represents the c-th reference signal matrix at the n-th time;

[0012] The error signal receiving unit obtains the error signal {e i (n)|i=1,2} and processed by the residual filter matrix, the two error filter signals {e hb (n)|b=1,2} and sent to the FELMS multi-channel noise reduction algorithm module, where e i (n) represents the i-th error signal at the n-th moment, e hb (n) represents the b-th error filtering signal at the n-th moment;

[0013] The FELMS multi-channel noise reduction algorithm module uses the identification filter matrix and the residual filter matrix to receive the two reference signal matrices {X c (n)|c=1,2} is processed to obtain the reference signal matrix {R hb (n)|b=1,2}, where R hb (n) represents the reference signal matrix after filtering and compensation at the bth time;

[0014] The FELMS multi-channel noise reduction algorithm module uses the LMS algorithm to filter the two error filtering signals {ehb (n)|b=1,2} and the reference signal matrix {R hb (n)|b=1,2} is processed to obtain the two filter weight coefficient increment matrices {ΔH b (n)|b=1,2}, where ΔH b (n) represents the incremental matrix of the b-th filter weight coefficient at the n-th moment;

[0015] The FELMS multi-channel noise reduction algorithm module is based on the two filter weight coefficient increment matrices {ΔH b (n)|b=1,2} and step size factor μ, for the four adaptive filter weight coefficient matrices {H cb (n)|c=1,2;b=1,2} are updated, and the updated 4 adaptive filter weight coefficients at the nth moment are updated to the matrix {H' cb (n)|c=1,2;b=1,2} and the two reference signal matrices {X c (n)|c=1,2} are convolved and summed to generate two anti-noise signals {y b (n)|b=1,2} and sent to the anti-noise signal sending unit through the audio data transmission module; wherein H cb (n) represents the adaptive filter weight coefficient matrix at the cth row and bth column at the nth time, H' cb (n) represents the adaptive filter weight coefficient update matrix at the cth row and bth column at the nth time, y b (n) represents the b-th anti-noise signal at the n-th moment;

[0016] The anti-noise signal sending unit sends two anti-noise signals {y b (n)|b=1,2} output, and compared with the noise signal {x i After superposition, the error signal {e i (n+1)|i=1,2} and transmit it to the error signal receiving unit, where e i (n+1) represents the i-th error signal at the n+1th moment, x i (n) represents the noise signal at the i-th target noise reduction point at the n-th time.

[0017] The present invention provides a multi-channel active noise reduction method for improving sound quality in a vehicle, comprising the following steps:

[0018] Step 1: Define the current time as n, the noise reduction step factor as μ, and the acquisition time as L;

[0019] Step 2: Obtain the noise source signal x(n) at the nth moment and perform spectrum analysis to obtain the frequency domain amplitude spectrum and energy spectrum information of the noise source signal x(n) at the nth moment;

[0020] Step 3: Use the sound quality parameter calculation model to process the frequency domain energy spectrum information to obtain the critical band energy and critical band sound pressure level of the noise source signal x(n) at the nth moment;

[0021] Step 4: Calculate the sound quality parameter eigenvalue matrix of the noise source signal x(n) at the nth moment according to the critical frequency band sound pressure level;

[0022] Step 5: Select the frequency values ​​and amplitudes corresponding to the two largest eigenvalues ​​from the sound quality parameter eigenvalue matrix based on the frequency domain amplitude spectrum and energy spectrum information;

[0023] Step 6: Based on the two frequency values ​​and their amplitudes, use formula (1) to generate two reference signal matrices {X c (n)|c=1,2}; where X c (n) represents the c-th reference signal matrix at the n-th time;

[0024]

[0025] In formula (1), x c (n) is the preset c-th reference signal at the n-th moment, x c (n-1) is the preset c-th reference signal at the n-1th moment, x c (n-L+1) is the cth reference signal at the preset n-L+1th time;

[0026] Step 7: Use the residual filter to filter the error signal {e i (n)|i=1,2} are processed to obtain the two error filtering signals {e hb (n)|b=1,2}; where e i (n) represents the error signal at the i-th target noise reduction point at the n-th time, e hb (n) represents the b-th error filtering signal at the n-th moment;

[0027] Step 8: Use formula (2) to calculate the two reference signal matrices {X c (n)|c=1,2} is processed to obtain the reference signal matrix {R hb (n)|b=1,2}, where R hb (n) represents the reference signal matrix after filtering and compensation at the bth time;

[0028] R hb (n) = X c T (n)H Scb (n)H nwcb (n) (2)

[0029] In formula (2), {H Scb (n)|c=1,2,b=1,2} represents the four identification filter matrices at the nth moment, {H nwcb (n)|c=1,2,b=1,2} represents the four residual filter matrices at the nth moment, H Scb (n) represents the identification filter matrix of row c and column b at the nth moment, H nwcb (n) represents the residual filter matrix of row c and column b at the nth moment;

[0030] Step 9: Use LMS algorithm to filter the two error signals {e hb (n)|b=1,2} and the reference signal matrix {R hb (n)|b=1,2} is processed to obtain the two weight coefficient increment matrices {ΔH b (n)|b=1,2}, where ΔH b (n) represents the b-th weight coefficient increment matrix at the n-th moment;

[0031] Step 10: Use formula (3) to set the adaptive filter weight coefficient matrix {H cb (n)|c=1,2;b=1,2} are updated to obtain the updated matrix {H' cb (n)|c=1,2;b=1,2}, where H cb (n) represents the adaptive filter weight coefficient matrix at the cth row and bth column at the nth time, H' cb (n) represents the adaptive filter weight coefficient update matrix at the cth row and bth column at the nth time;

[0032] H' cb (n) = H cb (n)-μΔH b (n) (3)

[0033] Step 11: Generate two anti-noise signals {y b (n)|b=1,2}, where y b (n) represents the b-th anti-noise signal at the n-th moment;

[0034] y b (n) = X cT (n)*H' cb (n) (4)

[0035] In formula (4), * represents convolution summation;

[0036] Step 12: The two anti-noise signals {y b (n)|b=1,2} and the noise signal {x i After superposition, the error signal {e i (n+1)|i=1,2}, where x i (n) represents the noise signal at the i-th target noise reduction point at the n-th moment, e i (n+1) represents the error signal at the i-th target noise reduction point at the n+1th time;

[0037] Step 13: After assigning n+1 to n, return to step 2 and execute sequentially.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] 1. To improve the sound quality of in-vehicle noise, the present invention adds a noise signal analysis and preprocessing module to the traditional active noise control system, which includes time-to-frequency data conversion of noise source noise signals and calculation and analysis of noise sound quality parameters. This provides more accurate analysis of the sound quality characteristics of noise sources generating in-vehicle noise, facilitates the generation of corresponding reference signal frequency and amplitude information during the noise reduction process, and makes the controlled target noise more targeted and effective.

[0040] 2. This invention utilizes the FELMS multi-channel noise reduction algorithm module to design the residual filter matrix using an A-weighting curve that better reflects the human ear's subjective perception of sound levels at different frequencies. This improves in-vehicle sound quality by taking the human ear's subjective perception into account. Furthermore, a multi-channel noise reduction method that supports multiple reference signals improves interior noise reduction requirements while enhancing passenger comfort. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the working process of the system of the present invention;

[0042] Figure 2 This is a flow chart of the noise signal analysis preprocessing module in the present invention;

[0043] Figure 3 It is a signal flow diagram of the FELMS multi-channel noise reduction algorithm module in the present invention. DETAILED DESCRIPTION

[0044] In this embodiment, if Figure 1 As shown, a multi-channel active noise reduction system for improving vehicle sound quality includes: an audio transceiver module, an audio data transmission module, a noise signal analysis and preprocessing module, a reference signal generation module, a FELMS-based multi-channel noise reduction algorithm module, and a basic parameter setting module; wherein the audio transceiver module includes: a reference signal receiving unit, an error signal receiving unit, and an anti-noise signal sending unit; the noise signal analysis and preprocessing module includes: a time / frequency data conversion unit and a sound quality parameter calculation and analysis unit;

[0045] The reference microphone in the reference signal receiving unit is arranged near noise sources such as the engine in the vehicle; the two error microphones in the error signal receiving unit are arranged on the right side of the driver's headrest and the left side of the co-driver's headrest respectively; the two speakers in the anti-noise signal sending unit are arranged above the backrest of the rear seat in the vehicle, horizontally opposite to the two error microphones; the audio data transmission module adopts the AN831 audio module; the noise signal analysis and preprocessing module, the reference signal generation module, the FELMS-based multi-channel noise reduction algorithm module and the basic parameter setting module are uniformly integrated on the FPGA development board, and the overall system design is realized by combining board-level storage and logical operation resources.

[0046] Considering the system resource occupation, system control accuracy and timing design requirements, the signal sampling frequency is set to 8kHz, the noise reduction step factor μ=-0.000895 based on the FELMS multi-channel noise reduction algorithm module and the data acquisition time L=256 in the reference signal generation module, the four adaptive filter weight coefficient matrices {H cb (n)|c=1,2;b=1,2}, used for data initialization, where H cb (n) represents the adaptive filter weight coefficient matrix at the cth row and bth column at the nth moment;

[0047] The reference signal receiving unit obtains the noise source signal x(n) at the nth moment through the reference microphone and sends it to the time / frequency data conversion unit through the audio data transmission module;

[0048] like Figure 2 As shown, the time / frequency data conversion unit uses a fast Fourier transform method (such as a radix 4 DIT-FFT algorithm) to perform spectrum analysis on the noise source signal x(n) at the nth moment, obtains the specific signal amplitude and energy value corresponding to the frequency contained in the frequency domain amplitude spectrum and energy spectrum information of the noise source signal x(n) at the nth moment, and sends it to the sound quality parameter calculation and analysis unit;

[0049] The sound quality parameter calculation and analysis unit uses a sound quality parameter calculation model to process the specific signal amplitude and energy value corresponding to the frequency contained in the frequency domain amplitude spectrum and energy spectrum information, and divides the critical frequency band using the human ear auditory characteristics and the 1 / 3 octave frequency band division method to obtain the critical frequency band energy and sound pressure level of the noise source signal x(n) at the nth moment, thereby calculating the sound quality parameter eigenvalue matrix of the noise source signal x(n) at the nth moment based on the critical frequency band sound pressure level, and selecting the frequency values ​​and amplitudes corresponding to the two largest eigenvalues ​​from the sound quality parameter eigenvalue matrix based on the frequency domain amplitude spectrum and energy spectrum information, and then sending them to the reference signal generation module to set the internal frequency value and corresponding amplitude signal of the reference signal generation module;

[0050] The reference signal generation module stores the internal data and sets the signal according to the two frequency values ​​and their amplitudes, and the acquisition time L, and generates two reference signal matrices {X c (n)|c=1,2} and sent to the FELMS multi-channel noise reduction algorithm module, where X c (n) represents the c-th reference signal matrix at the n-th time;

[0051] The error signal receiving unit obtains the error signal {e at the two target noise reduction points at the nth moment through the error microphone i (n)|i=1,2} and processed by the residual filter matrix, the two error filter signals {e hb (n)|b=1,2} and sent to the FELMS multi-channel noise reduction algorithm module, where e i (n) represents the i-th error signal at the n-th moment, e hb (n) represents the b-th error filtering signal at the n-th moment;

[0052] like Figure 3 As shown, the FELMS multi-channel noise reduction algorithm module uses the identification filter matrix obtained by identifying the secondary physical channel and the residual filter matrix designed by fitting the A-weighted curve that can better reflect the subjective perception characteristics of the human ear for the sound size of different frequencies to receive the two reference signal matrices {X c (n)|c=1,2} is processed to obtain the reference signal matrix {R hb (n)|b=1,2}, where R hb (n) represents the reference signal matrix after filtering and compensation at the bth time;

[0053] Based on the FELMS multi-channel noise reduction algorithm module, the LMS algorithm is used to filter the two error filtering signals {e hb(n)|b=1,2} and the reference signal matrix {R hb (n)|b=1,2} is processed to obtain the two filter weight coefficient increment matrices {ΔH b (n)|b=1,2}, where ΔH b (n) represents the incremental matrix of the b-th filter weight coefficient at the n-th moment;

[0054] Based on the FELMS multi-channel noise reduction algorithm module, the weight coefficient increment matrix {ΔH b (n)|b=1,2} and step size factor μ, the four adaptive filter weight coefficient matrices {H cb (n)|c=1,2;b=1,2} are updated, and the updated 4 adaptive filter weight coefficients at the nth moment are updated to the matrix {H' cb (n)|c=1,2;b=1,2} and the two reference signal matrices {X c (n)|c=1,2} performs convolution and summation, specifically multiplication and addition operations on data, thereby generating two anti-noise signals {y b (n)|b=1,2} and sent to the anti-noise signal sending unit through the audio data transmission module; wherein H cb (n) represents the adaptive filter weight coefficient matrix at the cth row and bth column at the nth time, H' cb (n) represents the adaptive filter weight coefficient update matrix at the cth row and bth column at the nth time, y b (n) represents the b-th anti-noise signal at the n-th moment;

[0055] The anti-noise signal sending unit sends the two anti-noise signals {y b (n)|b=1,2} is output through the speaker and is combined with the noise signal {x i After superposition, the error signal {e i (n+1)|i=1,2} and transmit it to the error signal receiving unit, where e i (n+1) represents the i-th error signal at the n+1th moment, x i (n) represents the noise signal at the i-th target noise reduction point at the n-th time.

[0056] This system can achieve stable control of target noise during continuous operation and keep the error signal at the noise reduction target point at a low level, so that it can improve the sound quality inside the vehicle while meeting the requirements of spatial noise reduction.

[0057] In this embodiment, a multi-channel active noise reduction method for improving sound quality in a vehicle includes the following steps:

[0058] Step 1: Define the current time as n, define the noise reduction step factor as μ = -0.000895 and the acquisition time as L = 255;

[0059] Step 2: Obtain the noise source (e.g., engine) signal x(n) at the nth moment and perform spectrum analysis to obtain the frequency domain amplitude spectrum and energy spectrum information of the noise source signal x(n) at the nth moment;

[0060] Step 3: Use the sound quality parameter calculation model to process the frequency domain energy spectrum information to obtain the critical band energy and critical band sound pressure level of the noise source signal x(n) at the nth moment;

[0061] Step 4: Calculate the sound quality parameter eigenvalue matrix of the noise source signal x(n) at the nth moment according to the critical frequency band sound pressure level;

[0062] Step 5: Select the frequency values ​​and amplitudes corresponding to the two largest eigenvalues ​​from the sound quality parameter eigenvalue matrix based on the frequency domain amplitude spectrum and energy spectrum information;

[0063] Step 6: Based on the two frequency values ​​and their amplitudes, use formula (1) to generate two reference signal matrices {X c (n)|c=1,2}; where X c (n) represents the c-th reference signal matrix at the n-th time;

[0064]

[0065] In formula (1), x c (n) is the preset c-th reference signal at the n-th moment, x c (n-1) is the preset c-th reference signal at the n-1th moment, x c (n-L+1) is the cth reference signal at the preset n-L+1th time;

[0066] Step 7: Use the residual filter to filter the error signal {e i (n)|i=1,2} are processed to obtain the two error filtering signals {e hb (n)|b=1,2}; where e i (n) represents the error signal at the i-th target noise reduction point at the n-th time, e hb (n) represents the b-th error filtering signal at the n-th moment;

[0067] Step 8: Use formula (2) to calculate the two reference signal matrices {X c(n)|c=1,2} is processed to obtain the reference signal matrix {R hb (n)|b=1,2}, where R hb (n) represents the reference signal matrix after filtering and compensation at the bth time;

[0068] R hb (n) = X c T (n)H Scb (n)H nwcb (n) (2)

[0069] In formula (2), {H Scb (n)|c=1,2,b=1,2} represents the four identification filter matrices at the nth moment, {H nwcb (n)|c=1,2,b=1,2} represents the four residual filter matrices at the nth moment, H Scb (n) represents the identification filter matrix of row c and column b at the nth moment, H nwcb (n) represents the residual filter matrix of row c and column b at the nth moment;

[0070] Step 9: Use LMS algorithm to filter the two error signals {e hb (n)|b=1,2} and the reference signal matrix {R hb (n)|b=1,2} is processed to obtain the two weight coefficient increment matrices {ΔH b (n)|b=1,2}, where ΔH b (n) represents the b-th weight coefficient increment matrix at the n-th moment;

[0071] Step 10: Use formula (3) to set the adaptive filter weight coefficient matrix {H cb (n)|c=1,2;b=1,2} are updated to obtain the updated matrix {H' cb (n)|c=1,2;b=1,2}, where H cb (n) represents the adaptive filter weight coefficient matrix at the cth row and bth column at the nth time, H' cb (n) represents the adaptive filter weight coefficient update matrix at the cth row and bth column at the nth time;

[0072] H' cb (n) = H cb (n)-μΔH b (n) (3)

[0073] Step 11: Generate two anti-noise signals {yb (n)|b=1,2}, where y b (n) represents the b-th anti-noise signal at the n-th moment;

[0074] y b (n) = X c T (n)*H' cb (n) (4)

[0075] In formula (4), * represents convolution summation;

[0076] Step 12: The two anti-noise signals {y b (n)|b=1,2} and the noise signal {x i After superposition, the error signal {e i (n+1)|i=1,2}, where x i (n) represents the noise signal at the i-th target noise reduction point at the n-th moment, e i (n+1) represents the error signal at the i-th target noise reduction point at the n+1th time;

[0077] Step 13: After assigning n+1 to n, return to step 2 and execute sequentially.

Claims

1. A multi-channel active noise reduction method for improving sound quality in a vehicle, characterized by: The steps include: Step 1: Define the current time as n and the noise reduction step factor as and the collection duration is L; Step 2: Obtain the noise source signal x(n) at the nth moment and perform spectrum analysis to obtain the frequency domain amplitude spectrum and energy spectrum information of the noise source signal x(n) at the nth moment; Step 3: Use the sound quality parameter calculation model to process the frequency domain energy spectrum information to obtain the critical band energy and critical band sound pressure level of the noise source signal x(n) at the nth moment; Step 4: Calculate the sound quality parameter eigenvalue matrix of the noise source signal x(n) at the nth moment according to the critical frequency band sound pressure level; Step 5: Select the frequency values ​​and amplitudes corresponding to the two largest eigenvalues ​​from the sound quality parameter eigenvalue matrix based on the frequency domain amplitude spectrum and energy spectrum information; Step 6: Based on the two frequency values ​​and their amplitudes, use formula (1) to generate two reference signal matrices {X c (n)|c=1,2}; where X c (n) represents the c-th reference signal matrix at the n-th time; (1) In formula (1), is the pre-set c-th reference signal at the n-th moment, is the cth reference signal at the preset time n-1, is the cth reference signal at the preset time n-L+1; Step 7: Use the residual filter to filter the error signal {e i (n)|i=1,2} are processed to obtain the two error filtering signals {e hb (n)|b=1,2}; where e i (n) represents the error signal at the i-th target noise reduction point at the n-th time, e hb (n) represents the b-th error filtering signal at the n-th moment; Step 8: Use formula (2) to calculate the two reference signal matrices {X c (n)|c=1,2} is processed to obtain the reference signal matrix {R hb (n)|b=1,2}, where R hb (n) represents the reference signal matrix after filtering and compensation at the bth time; (2) In formula (2), {H Scb (n)|c=1,2,b=1,2} represents the four identification filter matrices at the nth moment, {H nwcb (n)|c=1,2,b=1,2} represents the four residual filter matrices at the nth moment, H Scb (n) represents the identification filter matrix of row c and column b at the nth moment, H nwcb (n) represents the residual filter matrix of row c and column b at the nth moment; Step 9: Use LMS algorithm to filter the two error signals {e hb (n)|b=1,2} and the reference signal matrix {R hb (n)|b=1,2} is processed to obtain the two weight coefficient increment matrices {ΔH b (n)|b=1,2}, where ΔH b (n) represents the b-th weight coefficient increment matrix at the n-th moment; Step 10: Use formula (3) to set the adaptive filter weight coefficient matrix {H cb (n)|c=1,2;b=1,2} are updated to obtain the updated matrix {H' cb (n)|c=1,2;b=1,2}, where H cb (n) represents the adaptive filter weight coefficient matrix at the cth row and bth column at the nth time, H' cb (n) represents the adaptive filter weight coefficient update matrix at the cth row and bth column at the nth time; (3) Step 11: Generate two anti-noise signals {y b (n)|b=1,2}, where y b (n) represents the b-th anti-noise signal at the n-th moment; (4) In formula (4), represents convolution summation; Step 12: The two anti-noise signals {y b (n)|b=1,2} and the noise signal {x i (n)|i=1,2} are superimposed to generate the error signal {e i (n+1)|i=1,2}, where x i (n) represents the noise signal at the i-th target noise reduction point at the n-th moment, e i (n+1) represents the error signal at the i-th target noise reduction point at the n+1th time; Step 13: After assigning n+1 to n, return to step 2 and execute sequentially.

2. A multi-channel active noise reduction system for improving vehicle sound quality, using the multi-channel active noise reduction method for improving vehicle sound quality as claimed in claim 1, characterized in that: The in-vehicle multi-channel active noise reduction system includes: an audio transceiver module, an audio data transmission module, a noise signal analysis and preprocessing module, a reference signal generation module, a FELMS-based multi-channel noise reduction algorithm module, and a basic parameter setting module; wherein the audio transceiver module includes: a reference signal receiving unit, an error signal receiving unit, and an anti-noise signal sending unit; the noise signal analysis and preprocessing module includes: a time / frequency data conversion unit and a sound quality parameter calculation and analysis unit; The step size factor is set in the basic parameter setting module And the acquisition time L, the four adaptive filter weight coefficient matrices {H cb (n)|c=1,2;b=1,2}, where H cb (n) represents the adaptive filter weight coefficient matrix at the cth row and bth column at the nth moment; The reference signal receiving unit obtains the noise source signal x(n) at the nth moment and sends it to the time / frequency data conversion unit through the audio data transmission module; The time / frequency data conversion unit performs spectrum analysis on the noise source signal x(n) at the nth moment using a fast Fourier transform method to obtain frequency domain amplitude spectrum and energy spectrum information of the noise source signal x(n) at the nth moment and sends the information to the sound quality parameter calculation and analysis unit; The sound quality parameter calculation and analysis unit processes the frequency domain amplitude spectrum and energy spectrum information using the sound quality parameter calculation model to obtain the critical frequency band energy and sound pressure level of the noise source signal x(n) at the nth moment, thereby calculating the sound quality parameter eigenvalue matrix of the noise source signal x(n) at the nth moment according to the sound pressure level, and selecting the frequency values ​​and amplitudes corresponding to the two largest eigenvalues ​​from the sound quality parameter eigenvalue matrix based on the frequency domain amplitude spectrum and energy spectrum information and sending them to the reference signal generation module; The reference signal generation module generates two reference signal matrices {X c (n)|c=1,2} and sent to the FELMS multi-channel noise reduction algorithm module, where X c (n) represents the c-th reference signal matrix at the n-th time; The error signal receiving unit obtains the error signal {e i (n)|i=1,2} and processed by the residual filter matrix, the two error filter signals {e hb (n)|b=1,2} and sent to the FELMS multi-channel noise reduction algorithm module, where e i (n) represents the i-th error signal at the n-th moment, e hb (n) represents the b-th error filtering signal at the n-th moment; The FELMS multi-channel noise reduction algorithm module uses the identification filter matrix and the residual filter matrix to receive the two reference signal matrices {X c (n)|c=1,2} is processed to obtain the reference signal matrix {R hb (n)|b=1,2}, where R hb (n) represents the reference signal matrix after filtering and compensation at the bth time; The FELMS multi-channel noise reduction algorithm module uses the LMS algorithm to filter the two error filtering signals {e hb (n)|b=1,2} and the reference signal matrix {R hb (n)|b=1,2} is processed to obtain the two filter weight coefficient increment matrices {ΔH b (n)|b=1,2}, where ΔH b (n) represents the incremental matrix of the b-th filter weight coefficient at the n-th moment; The FELMS multi-channel noise reduction algorithm module is based on the two filter weight coefficient increment matrices {ΔH b (n)|b=1,2} and step size factor , for the four adaptive filter weight coefficient matrices {H cb (n)|c=1,2;b=1,2} are updated, and the updated 4 adaptive filter weight coefficients at the nth moment are updated to the matrix {H' cb (n)|c=1,2;b=1,2} and the two reference signal matrices {X c (n)|c=1,2} are convolved and summed to generate two anti-noise signals {y b (n)|b=1,2} and sent to the anti-noise signal sending unit through the audio data transmission module; wherein H cb (n) represents the adaptive filter weight coefficient matrix at the cth row and bth column at the nth time, H' cb (n) represents the adaptive filter weight coefficient update matrix at the cth row and bth column at the nth time, y b (n) represents the b-th anti-noise signal at the n-th moment; The anti-noise signal sending unit sends two anti-noise signals {y b (n)|b=1,2} output, and compared with the noise signal {x i (n)|i=1,2} are superimposed to generate the error signal {e i (n+1)|i=1,2} and transmit it to the error signal receiving unit, where e i (n+1) represents the i-th error signal at the n+1th moment, x i (n) represents the noise signal at the i-th target noise reduction point at the n-th time.

Images