A Design Method of Active Noise Control System Based on FBFULMS Algorithm

Abstract

The present invention claims protection for a design method for an active noise control system based on the FBFULMS algorithm. It mainly includes three parts: (1) NFBLMS filter design (2) MFBLMS filter design (3) FBFULMS filter design. The innovation of the present invention is that the frequency domain block filter U least mean square (FBFULMS) algorithm structure is adopted to ensure the stability and fast convergence speed of the system through secondary path modeling. The results show that when the input excitation is the compressor noise data, the proposed FBFULMS algorithm achieves a noise attenuation of up to 26dBA through 39 iterative steps. Finally, XILINX Artix7‑100TFPGA is used as the core control module for hardware design. The measurement results show that the method effectively attenuates the low-frequency and medium-frequency noise in the target area of ​​the compressor, especially at the error microphone, the measurement noise is reduced by up to 20dBA, and the convergence time is about 60ms.

Description

Technical Field

[0001] The invention belongs to the field of digital signal processing, and in particular relates to an active noise control system based on a FBFULMS algorithm and a design method thereof. Background Art

[0002] Currently, Active Noise Control (ANC) is widely used in automobiles, headphones, and aerospace fields for its good low-frequency noise elimination performance. It is different from dampers, barriers, mufflers, or acoustic crystals that are only suitable for high-frequency noise. For general feedforward control systems, the Filtered-X Least Mean Square (FxLMS) algorithm is widely used due to its stability and simplicity.

[0003] Many researchers have pointed out that the time-domain FXLMS algorithm has a slow convergence speed and a long filter length. To address these difficulties, many effective solutions have been proposed, such as the frequency-domain block least mean square (FBLMS) algorithm. The NFBLMS algorithm is obtained by normalizing the list range of the FBLMS algorithm in each frequency band with respect to the corresponding reference signal power, which has a consistent convergence pattern within each frequency band. Although the NFBLMS algorithm has a fast convergence speed for colored noise, when the adaptive filter length is insufficient, its mean square error (MSE) is large due to the existence of a secondary path between the control source and the error microphone.

[0004] This is a common situation in ANC systems. For these situations of NFBLMS, researchers proposed the MFBLMS algorithm, but the convergence speed of the MFBLMS algorithm is usually lower than that of the NFBLMS algorithm. In order to address these limitations of the NFBLMS and MFBLMS algorithms, a hybrid algorithm with a dynamic switching strategy between the MFBLMS and NFBLMS algorithms is proposed, which combines the advantages of NFBLMS and MFBLMS. However, it still cannot achieve a good state in terms of noise reduction amount and convergence speed. Then a study on active noise control system based on the FBFULMS algorithm is proposed.

[0005] Patent CN110808025A, a modular design method for an active noise control system based on a field programmable gate array (FPGA). It mainly includes 6 modules: a white noise signal generator, a convolution module 3, a weight update module 2, a convolution module 2, a weight update module 1, and a convolution module 1. This invention proposes an active noise control system based on the momentum least mean square (MLMS) algorithm, the purpose of which is to improve the operation speed of the ANC system and the system's data throughput. The solution proposed by patent CN110808025A improves the operation speed of the ANC system and the system's data throughput, but the system's stability is poor.

[0006] The present invention proposes a design method for an active noise control system based on the FBFULMS algorithm. The innovation of the present invention is that the frequency domain block filter U least mean square (FBFULMS) algorithm structure is adopted to improve the operation speed of the system through secondary path modeling, while ensuring the stability of the system and the actual noise reduction effect of the algorithm. The proposed FBFULMS algorithm is verified on the FPGA board. The results show that when the input excitation is the compressor noise data, after 39 iterations, the noise attenuation of up to 26dBA is achieved, especially at the error microphone, the measurement noise is reduced by up to 20dBA, and the convergence time is about 60ms. Summary of the invention

[0007] The present invention aims to solve the above problems of the prior art. A design method of an active noise control system based on the FBFULMS algorithm is proposed. The technical solution of the present invention is as follows:

[0008] A design method for an active noise control system based on the FBFULMS algorithm comprises the following steps:

[0009] (1) Design steps of the normalized frequency-domain block least-mean-square (NFBLMS) filter: It is essentially a fast implementation of the BLMS (block least-mean-square) algorithm. It uses the overlap-preserve method to process the input signal of the FIR filter in blocks, uses circular convolution to implement linear convolution, and then uses the circular convolution principle of FFT to implement time-domain circular convolution by frequency-domain multiplication and adaptively adjust the frequency-domain filter coefficients, and then normalizes the convergence factor to achieve adaptive filtering;

[0010] (2) Design steps of improved frequency-domain least mean square algorithm MFBLMS filter: The improved frequency-domain block least mean square algorithm (MFBLMS) adaptive filter is essentially a modified frequency-domain block least mean square algorithm proposed because the average weight vector of the normalized frequency-domain block least mean square (NFBLMS) algorithm cannot converge to the optimal solution in terms of mean square error.

[0011] (3) Design steps of frequency domain block filter U least mean square FBFULMS filter: (secondary path modeling is performed based on the hybrid algorithm of MFBLMS and NFBLMS).

[0012] Furthermore, the binary normalized frequency domain least mean square NFBLMS filter specifically includes:

[0013] The first FBLMS fast frequency domain adaptive algorithm module is used for a fast (numerically efficient) implementation of the BLMS (block least-mean-square, BLMS) algorithm in the frequency domain.

[0014] The first weight update module is used to update N weight coefficients. The NFBLMS filter weight update iterative calculation formula is:

[0015]

[0016] The superscript H represents the conjugate transpose operation; μ is the step factor, normalized to C, is the weight coefficient in the frequency domain, is the constraint matrix, is the mean square error, and m represents the number of filter input signals X. f (m) = diag[x f (m)],M f =diag[ξ], so ξ is usually set to the inverse of the reference signal power Q N,0 =FG N,0 F -1 ,Q 0,N =FG 0,N F -1 , represents a 2N*2N matrix, N represents the block length of BLMS, and F represents the FFT operation. -1 Indicates IFFT operation.

[0017] Furthermore, the filter coefficient update equation of the frequency domain least mean square algorithm MFBLMS algorithm is:

[0018]

[0019] That is, the weight update coefficient is

[0020] When the adaptive filter length is insufficient, the steady-state solution of the NFBLMS algorithm deviates from the optimal Wiener solution, while the MFBLMS algorithm still converges to the Wiener solution.

[0021] Furthermore, the filter frequency domain coefficients of the MFBLMS adaptive filter are:

[0022]

[0023] Where 0 is an M*1 zero sequence, FFT[] is a fast Fourier transform operation on the sequence, and the length of the frequency domain coefficient W(k) here is twice the length of the time domain coefficient w(k);

[0024] Let the (frequency domain) input signal be:

[0025] Where u() represents the input signal in the time domain. This formula represents the 2M*2M diagonal matrix obtained by Fourier transforming the continuous input data sub-blocks. For the overlap-preserve method, linear convolution can be implemented by circular convolution to obtain an M*1 sequence:

[0026] y Τ =[y(kM),y(kM+1),...,y(kM+M-1)]=IFFT[U(k)W(k)] the last M elements.

[0027] Where y() is the output signal in the frequency domain, and IFFT[] is the inverse Fourier transform of the sequence;

[0028] The expected response sequence in the frequency domain is:

[0029] d=[d(kM),d(kM+1),...,d(kM+M-1)] T

[0030] The frequency domain error signal is:

[0031]

[0032] Where e(k) is the error sequence in the time domain.

[0033] Furthermore, the BLMS filter (which is the time-domain part of the FBLMS filter) divides the input signal u(n) into blocks of length L through a serial-to-parallel converter, and the resulting input data blocks are successively added to a FIR filter of length M; after collecting the data values ​​of each block, the filter coefficients are updated, so that the adaptive update of the filter is performed in a block operation manner, rather than point by point like a time-domain LMS filter;

[0034] Its input signal is:

[0035] u(n)=[u(n),u(n-1),...,u(n-M+1)] T

[0036] represents the input signal sequence at time n, T represents the transposition, and accordingly, let:

[0037] W(n)=[w0(n),w1(n),...,w(n-M+1)] T

[0038] represents the filter coefficient at time n, k is the block index, and its relationship with n is:

[0039] n=kL+i,i=0,1,2...,L-1,k=1,2...

[0040] Where L is the block length, and the input signal of the kth block is defined as Its matrix form is:

[0041] A T (k)=[u(kL),u(kL+1),...,u(kL+L-1)]

[0042] During the duration of the input data block, the filter coefficient remains W(k), which is the value of W(k) when n=k.

[0043] Furthermore, the hybrid NFBLMS filter is a switching improvement of the advantages of the NFBLMS and MFBLMS algorithms, and the specific implementation steps are as follows:

[0044] Step (1) initializing the algorithm using the NFBLMS algorithm, and when the NFBLMS algorithm should reach a stable state, switching the proposed hybrid MFBLMS to the MFBLMS algorithm;

[0045] Step (2) When the MFBLMS algorithm converges to an approximation of the Wiener solution, it approaches the minimum;

[0046] Step (3) After the MFBLMS algorithm converges, the system is converted back to the NFBLMS algorithm to naturally obtain a steady-state value, which is then used as a threshold;

[0047] Step (4) converting the NFBLMS algorithm into the MFBLMS algorithm;

[0048] Step (5) When a significant change in the acoustic environment is detected, the MFBLMS algorithm is converted to the NFBLMS algorithm which converges faster.

[0049] Furthermore, the weight coefficient update equation of the FBFULMS filter is improved as follows:

[0050]

[0051] The weight coefficient update equation for its secondary path modeling is:

[0052]

[0053] The superscript H represents the conjugate transpose operation, and the step factors μ1 and μ2 are normalized to ξ1 and ξ2. To ensure the stability of the entire system, ξ=min(ξ1,ξ2), X f (m) = diag[x f (m)], the formula is improved to:

[0054]

[0055] P f is the power matrix composed of the desired signal and the input signal in the time domain, O f is the power matrix composed of the desired signal and the input signal in the frequency domain, N f is the power matrix composed of the desired signal in the frequency domain, M f is the power matrix composed of the input signal in the frequency domain.

[0056] In order to measure the convergence performance of the system, the minimum mean square error can be expressed as:

[0057]

[0058] By calculating the gradient of the above formula, we can get:

[0059]

[0060] By setting the gradient function of the above formula to 0, the optimal weight vector and can be derived:

[0061]

[0062] Therefore, the global minimum of the minimum mean square error can be achieved. At the same time, the error signal in the frequency domain can be described as:

[0063] e f (m) = FG 0,N F -1 [d f (m)-X f (k)(a f (m)+b f (m))]s.

[0064] Furthermore, the non-causal part of the filter coefficients in the error signal of the FBFULMS algorithm is not affected by the update process, so both sides are multiplied by the yield:

[0065]

[0066] And in the FBFULMS algorithm a pair of FFT / IFFT is added in each block.

[0067] The advantages and beneficial effects of the present invention are as follows:

[0068] The active noise control system design method based on the FBFULMS algorithm described in the present invention (1) is characterized in that the BLMS filter divides the input signal u(n) into blocks of length L through a serial-to-parallel converter, and the generated input data blocks are successively added to the FIR filter of length M. After collecting the data values ​​of each block, the filter coefficients are updated, so that the adaptive update of the filter is performed in a block operation mode, thereby improving the operation speed of the system.

[0069] (2) The hybrid NFBLMS filter described above, although the NFBLMS algorithm has a fast convergence speed for noise, when the adaptive filter length is insufficient, its mean square error (MSE) is large due to the existence of a secondary path between the control source and the error microphone. For these situations of NFBLMS, researchers have proposed the MFBLMS algorithm, but the convergence speed of the MFBLMS algorithm is usually lower than that of the NFBLMS algorithm. In order to address these limitations of the NFBLMS and MFBLMS algorithms, a hybrid algorithm is proposed, which has a dynamic switching strategy between the MFBLMS and NFBLMS algorithms. When the NFBLMS algorithm should reach a stable state, the proposed hybrid MFBLMS is switched to the MFBLMS algorithm; when the MFBLMS algorithm converges to an approximation of the Wiener solution, the system is converted back to the NFBLMS algorithm, and a steady-state value is naturally obtained, which is then used as a threshold; when a large change in the acoustic environment is detected, the MFBLMS algorithm is converted to the faster-converging NFBLMS algorithm.

[0070] (3) The FBFULMS algorithm is proposed because although the hybrid algorithm effectively combines the best steady-state behavior of the MFBLMS algorithm and the high convergence speed of the NFBLMS algorithm, the convergence speed and noise reduction of the active noise control system still cannot meet the requirements. Then the FBFULMS algorithm is proposed, which models the secondary path based on NFBLMS, so that the system not only remains stable, but also improves its convergence speed by synthesizing the required signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a hybrid algorithm block diagram of NFBLMS and MFBLMS provided by a preferred embodiment of the present invention;

[0072] Figure 2It is a block diagram of the FUFBLMS system algorithm provided by the present invention;

[0073] Figure 3 The simulation results of the convergence of the four algorithms provided by the present invention; DETAILED DESCRIPTION

[0074] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0075] The technical solution of the present invention to solve the above technical problems is:

[0076] The invention discloses an active noise control system based on FBFULMS algorithm. The bin-Normalized Frequency Domain Block Least Mean Square (NFBLMS) algorithm has a partial steady-state solution when the adaptive filter length is insufficient. Then the frequency domain least mean square algorithm (MFBLMS) is improved, which is an adaptive hybrid algorithm composed of NFBLMS and MFBLMS. Because of its stability and convergence speed, the structure of the frequency domain block filter U least mean square (FBFULMS) algorithm is finally proposed.

[0077] As attached Figure 1 As shown, the present invention proposes an active noise control system based on the FBFULMS algorithm, characterized in that the NFBLMS filter specifically includes:

[0078] The first fast frequency domain adaptive algorithm (Frequency Domain Block Least Mean Square, FBLMS) module is a fast frequency domain solution of the BLMS algorithm.

[0079] The first fast frequency domain adaptive algorithm (Frequency Domain Block Least Mean Square, FBLMS) module is a fast frequency domain solution of the BLMS algorithm.

[0080] The first weight update module is used to update N weight coefficients. The NFBLMS filter weight update iterative calculation formula is:

[0081]

[0082] The superscript H indicates the conjugate transpose operation; μ is the step factor, normalized to C,

[0083] X f (m) = diag[x f(m)],M f =diag[ξ] and ξ is usually set to the inverse of the reference signal power, so Q N,0 =FG N,0 F -1 ,Q 0,N =FG 0,N F -1 , represents a 2N*2N matrix, where F represents FFT operation and F-1 represents IFFT operation

[0084] When the adaptive filter length is insufficient, the steady-state solution of the NFBLMS algorithm deviates from the optimal Wiener solution, while the MFBLMS algorithm still converges to the Wiener solution. Due to more conditions on the equivalent reference autocorrelation matrix, the MFBLMS algorithm converges slower than the NFBLMS algorithm.

[0085] Furthermore, the FBLMS adaptive filter is essentially a fast implementation of the BLMS algorithm, which utilizes the overlap-retention method to process the input signal of the FIR filter in blocks, utilizes circular convolution to implement linear convolution, and then utilizes the circular convolution principle of FFT to implement time domain circular convolution by frequency domain multiplication and adaptively adjust the frequency domain filter coefficients, thereby achieving the purpose of adaptive filtering.

[0086] The filter frequency domain coefficients are:

[0087]

[0088] Where 0 is an M*1 zero sequence, and FFT[] is a fast Fourier transform operation on the sequence. The length of the frequency domain coefficient W(k) here is twice the length of the time domain coefficient w(k).

[0089] Furthermore, the hybrid NFBLMS filter is a switch improvement of the advantages of NFBLMS and MFBLMS algorithms. The steps are as follows:

[0090] Step (1) Initialize the algorithm using the NFBLMS algorithm. When the NFBLMS algorithm should reach a stable state, the proposed algorithm is switched to the MFBLMS algorithm.

[0091] Step (2) When the MFBLMS algorithm converges to an approximation of the Wiener solution, it approaches the minimum.

[0092] Step (3) After the MFBLMS algorithm converges, the system is converted back to the NFBLMS algorithm to naturally obtain a steady-state value which is then used as the threshold.

[0093] Step (4) converts the algorithm into the MFBLMS algorithm.

[0094] Step (5) When a significant change in the acoustic environment is detected, the algorithm is converted to the NFBLMS algorithm with faster convergence.

[0095] As attached Figure 2 As shown, the present invention provides an active noise control system based on the FBFULMS algorithm. The FBFULMS algorithm is proposed because although the hybrid algorithm effectively combines the best steady-state behavior of the MFBLMS algorithm and the high convergence speed of the NFBLMS algorithm, the convergence speed and noise reduction of the active noise control system still cannot meet the requirements. Then the FBFULMS algorithm is proposed, which models the secondary path based on NFBLMS, so that the system not only remains stable, but also improves its convergence speed by synthesizing the required signal.

[0096] Furthermore, the weight coefficient update equation of the FBFULMS algorithm is improved as follows:

[0097]

[0098] The weight coefficient update equation for its secondary path modeling is:

[0099]

[0100] The superscript H represents the conjugate transpose operation, and the step factors μ1 and μ2 are normalized to ξ1 and ξ2. To ensure the stability of the entire system, ξ=min(ξ1,ξ2), X f (m) = diag[x f (m)]. The formula is improved to:

[0101]

[0102] In order to measure the convergence performance of the system, the minimum mean square error can be expressed as:

[0103]

[0104] By calculating the gradient of the above formula, we can get:

[0105]

[0106] By setting the gradient function of the above formula to 0, the optimal weight vector and can be derived:

[0107]

[0108] Therefore, the global minimum of the minimum mean square error can be achieved. At the same time, the error signal in the frequency domain can be described as:

[0109] e f (m) = FG 0,N F-1 [d f (m)-X f (k)(a f (m)+b f (m))]s

[0110] As attached Figure 3 , represents the comparison of the convergence of different designed filters. It can be seen that the convergence of the FUFBLMS proposed in this study is better than that of the proposed NFBLMS, MFBLMS, and hybrid MFBLMS.

[0111] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0112] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0113] The above embodiments should be understood to be only used to illustrate the present invention and not to limit the protection scope of the present invention. After reading the contents of the present invention, technicians can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A design method for an active noise control system based on the FBFULMS algorithm, characterized in that: The following steps are involved: (1) Design steps of the normalized frequency domain least mean square NFBLMS filter: It is essentially a fast implementation of the BLMS algorithm. It uses the overlap-retain method to process the input signal of the FIR filter in blocks, uses circular convolution to implement linear convolution, and then uses the circular convolution principle of FFT to implement time domain circular convolution by frequency domain multiplication and adaptively adjust the frequency domain filter coefficients, and then normalizes the convergence factor to achieve adaptive filtering; (2) Design steps of the improved frequency domain least mean square algorithm MFBLMS filter: The improved frequency domain least mean square algorithm MFBLMS adaptive filter is essentially a proposed improved frequency domain block least mean square algorithm because the average weight vector of the normalized frequency domain block least mean square NFBLMS algorithm cannot converge to the optimal solution in terms of mean square error; (3) Design steps of frequency domain block filter U least mean square FBFULMS filter: perform secondary path modeling based on the hybrid algorithm of MFBLMS and NFBLMS; The hybrid NFBLMS algorithm is a switch improvement of the advantages of the NFBLMS and MFBLMS algorithms. The specific implementation steps are as follows: Step (1) initializing the algorithm using the NFBLMS algorithm, and when the NFBLMS algorithm should reach a stable state, switching the proposed hybrid MFBLMS to the MFBLMS algorithm; Step (2) When the MFBLMS algorithm converges to an approximation of the Wiener solution, it approaches the minimum; Step (3) After the MFBLMS algorithm converges, the system is converted back to the NFBLMS algorithm to naturally obtain a steady-state value, which is then used as a threshold; Step (4) converting the NFBLMS algorithm into the MFBLMS algorithm; Step (5) When a significant change in the acoustic environment is detected, the MFBLMS algorithm is converted to the NFBLMS algorithm with faster convergence; The weight coefficient update equation of the FBFULMS filter is improved as follows: a f (m+1)=a f (m)+2μ1Q N,0 M f X f H (m)e f (m) The weight coefficient update equation for its secondary path modeling is: The superscript H represents the conjugate transpose operation, and the step factors μ1 and μ2 are normalized to ξ1 and ξ2. To ensure the stability of the entire system, ξ=min(ξ1,ξ2), X f (m) = diag[x f (m)], the formula is improved to: P f is the power matrix composed of the desired signal and the input signal in the time domain, O f is the power matrix composed of the expected signal and the input signal in the frequency domain, N f is the power matrix composed of the desired signal in the frequency domain, M f The power matrix composed of the input signal in the frequency domain. In order to measure the convergence performance of the system, the minimum mean square error can be expressed as: By calculating the gradient of the above formula, we can get: By setting the gradient function of the above formula to 0, the optimal weight vector and can be derived: Therefore, the global minimum of the minimum mean square error can be achieved. At the same time, the error signal in the frequency domain can be described as: e f (m)=FG 0,N F -1 [d f (m)-X f (k)(a) f (m)+b f (m))]s.

2. The design method of the active noise control system based on the FBFULMS algorithm according to claim 1 is characterized in that: The normalized frequency domain least mean square NFBLMS filter specifically includes: The first FBLMS fast frequency domain adaptive algorithm module is used for a fast numerically efficient implementation of the BLMS algorithm in the frequency domain; The first weight update module is used to update N weight coefficients. The NFBLMS filter weight update iterative calculation formula is: The superscript H indicates the conjugate transpose operation; μ is the step factor, normalized to C, W f is the weight coefficient in the frequency domain, Q N,0 is the constraint matrix, e f (m) is the mean square error, m represents the number of filter input signals, X f (m) = diag[x f (m)],M f =diag[ξ], so ξ is usually set to the inverse of the reference signal power, Q N,0 =FG N,0 F -1 ,Q 0,N =FG 0,N F -1 , represents a 2N*2N matrix, N represents the block length of BLMS, and F represents FFT operation. -1 Indicates IFFT operation.

3. The design method of active noise control system based on FBFULMS algorithm according to claim 1, characterized in that: The filter coefficient update equation of the frequency domain least mean square algorithm MFBLMS algorithm is: That is, the weight update coefficient is When the adaptive filter length is insufficient, the steady-state solution of the NFBLMS algorithm deviates from the optimal Wiener solution, while the MFBLMS algorithm still converges to the Wiener solution.

4. The design method of active noise control system based on FBFULMS algorithm according to claim 1, characterized in that: The filter frequency domain coefficients of the MFBLMS adaptive filter are: Where 0 is an M*1 zero sequence, FFT[] is a fast Fourier transform operation on the sequence, and the length of the frequency domain coefficient W(k) here is twice the length of the time domain coefficient w(k); Let the (frequency domain) input signal be: Where u() represents the input signal in the time domain. This formula represents the 2M*2M diagonal matrix obtained by Fourier transforming the continuous input data sub-blocks. For the overlap-preserve method, linear convolution can be implemented by circular convolution to obtain an M*1 sequence: y Τ =[y(kM),y(kM+1),...,y(kM+M-1)]=IFFT[U(k)W(k)] last M elements; Where y() is the output signal in the frequency domain, and IFFT[] is the inverse Fourier transform of the sequence; The expected response sequence in the frequency domain is: d=[d(kM),d(kM+1),...,d(kM+M-1)] T The frequency domain error signal is: Where e(k) is the error sequence in the time domain.

5. The design method of active noise control system based on FBFULMS algorithm according to claim 4 is characterized in that: The BLMS filter is the time-domain part of the FBLMS filter. The input signal u(n) is divided into blocks of length L by a serial-to-parallel converter, and the resulting input data blocks are successively added to the FIR filter of length M; after collecting the data values ​​of each block, the filter coefficients are updated, so that the adaptive update of the filter is performed in a block operation manner, rather than point by point like the time-domain LMS filter; Its input signal is: u(n)=[u(n),u(n-1),...,u(n-M+1)] T represents the input signal sequence at time n, T represents the transposition, and accordingly, let: W(n)=[w0(n),w1(n),...,w(n-M+1)] T represents the filter coefficient at time n, k is the block index, and its relationship with n is: n=kL+i,i=0,1,2...,L-1,k=1,2... Where L is the block length, and the input signal of the kth block is defined as Its matrix form is: A T (k)=[u(kL),u(kL+1),...,u(kL+L-1)] During the duration of the input data block, the filter coefficient remains W(k), which is the value of W(k) when n=k.

6. The design method of active noise control system based on FBFULMS algorithm according to claim 1, characterized in that: The non-causal part of the filter coefficients in the error signal of the FBFULMS algorithm is not affected by the update process, so multiplying both sides by the yield: And in the FBFULMS algorithm a pair of FFT / IFFT is added in each block.

Images