Adaptive pruning multi-reference noise reduction method, device, system and storage medium

Abstract

The present invention discloses an adaptive pruning multi-reference noise reduction method, device, system, and storage medium, belonging to the field of control algorithm technology. The method includes: using a pre-built filtered-x least mean square algorithm to set iteratively updated self-adjusting parameters for each reference signal; assigning weights to each reference signal based on the self-adjusting parameters; filtering the reference signals based on the weights to obtain filtered signals; and using the filtered signals to offset noise signals to achieve noise reduction. This method can reduce the influence of irrelevant signals and achieve stable noise reduction.

Description

Technical Field

[0001] The present invention relates to an adaptive pruning multi-reference noise reduction method, device, system and storage medium, and belongs to the technical field of control algorithms. Background Art

[0002] Low-frequency noise is widely present in daily life and work, and can cause physical and mental damage. In practical applications, there are usually multiple noise sources. Taking a car as an example, there is noise generated by the engine, noise caused by tire-ground friction, noise caused by chassis vibration, and wind noise during high-speed driving. Therefore, practical systems need to adopt a multi-channel system with multiple reference signals to achieve high noise reduction.

[0003] The selection of the reference signal directly impacts the effectiveness of noise reduction. The stronger the correlation between the reference signal and the error signal, the better the noise reduction. Furthermore, the noise sources in real-world environments can be time-varying. For example, at speeds of 50 km / h and below, interior noise primarily comes from the transmission and engine. Above 50 km / h, the friction between the tires and the road begins to increase with speed. When speeds exceed 120 km / h, wind noise becomes the primary source of interior noise. Under time-varying noise sources, each reference signal contributes differently to the system's noise reduction performance. Preselecting a reference signal without adjusting it can lead to system performance degradation or even divergence. Furthermore, the continuous computation of multiple reference signals consumes a significant portion of system resources. Therefore, adaptively pruning these multiple reference signals can improve system stability and reduce computational complexity.

[0004] In the existing multi-reference signal active noise reduction technology, the reference signal reshaping method based on the generalized sidelobe canceller is used to improve the correlation between the reference signal and the target noise. Solution A combines a multi-channel feedforward ANC system with noise source separation. When the ANC system uses two reference microphones and the reference microphones do not need to be arranged near the noise source, the proposed system can improve the noise reduction performance; Solution B is a multi-channel active noise control system based on the optimal reference microphone selector of the arrival time difference. The selector selects the reference microphone that meets the causal constraint based on the time difference. The algorithm adjusts the amplitude level of the reference signal to alleviate the system instability caused by the different dynamic characteristics of different reference signals. The traditional Filtered-x Least Mean Square (Filtered-x Least Mean Square) The FXLMS (Frequency Loss Square) algorithm is based on the LMS algorithm and takes into account the influence of the secondary path. Due to its simple calculation, strong stability and ability to adaptively track the environment, the noise reduction effect and convergence speed are improved. In the application of the noise control system algorithm, a single secondary source and a single error microphone of J reference signals are used. Each reference signal must pass through an adaptive filter to generate a secondary signal. Due to the difference between the reference signals, each adaptive filter has an optimal step size to control the convergence speed. However, in order to ensure the stability of the system, only the smallest of all optimal step sizes can be selected, resulting in the output of each adaptive filter converging at a different speed, resulting in channel dependence problems, slowing the overall convergence speed, and reducing the performance of the system. Before the algorithm works, it is necessary to collect all independent signals related to the noise to be controlled, and then calculate the correlation coefficient of the collected signals, and select the ones with high correlation as reference signals. Summary of the Invention

[0005] The object of the present invention is to provide an adaptive pruning multi-reference noise reduction method, device, system and storage medium, which can reduce the influence of irrelevant signals and achieve stable noise reduction.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides an adaptive pruning multi-reference denoising method, comprising:

[0008] Using the pre-built filter-x least mean square algorithm, the self-tuning parameters for each reference signal are set for iterative update;

[0009] assigning a weight to each of the reference signals according to the self-adjusting parameter;

[0010] filtering the reference signal according to the weight to obtain a filtered signal;

[0011] The filtered signal is used to offset the noise signal, thereby achieving noise reduction.

[0012] In combination with the first aspect, further, the expression of the filter-x least mean square algorithm is:

[0013]

[0014] Among them, i is the i-th noise source, n is the n-th moment, x i (n) is the reference signal picked up by the sensor at the i-th noise source at the n-th moment, w i (n) is the filter at the nth moment, w i (n+1) is the filter at the n+1th moment, e(n) is the noise residual measured by the error microphone at the nth moment, s(n) is the impulse response of the secondary path at the nth moment, is the estimated value of the impulse response s(n) of the secondary path at the nth moment, λ i (n) is the reference signal x picked up at the i-th noise source at the n-th moment i (n) is the self-adjusting parameter, μ is the step size, and * is the linear convolution operation.

[0015] In combination with the first aspect, further, the expression of the self-adjusting parameter is:

[0016]

[0017] Among them, λ i (n) is the reference signal x picked up at the i-th noise source at the n-th moment i (n) self-adjusting parameter, a i (n) is the self-adjusting parameter variable at the nth moment.

[0018] In combination with the first aspect, further, the iterative update formula of the self-adjusting parameter is:

[0019]

[0020] Among them, a i (n) is the self-adjusting parameter variable at the nth moment, a i (n+1) is the self-adjusting parameter variable at the n+1th moment, μ a is the self-adjusting parameter variable a at the nth moment i (n) is the step size of the update, e(n) is the noise residual measured by the error microphone at the nth moment, w i (n) is the filter at the nth moment, x i (n) is the reference signal picked up by the sensor at the i-th noise source at the n-th moment, λ i (n) is the reference signal x picked up at the i-th noise source at the n-th moment i(n), s(n) is the self-adjusting parameter of the secondary path at the nth moment, and * is the linear convolution operation.

[0021] In combination with the first aspect, further, each of the filters adopts a finite impulse response filter with a transverse structure, and the order of the finite impulse response filter with a transverse structure is M.

[0022] In a second aspect, the present invention provides an adaptive pruning multi-reference noise reduction device, comprising:

[0023] Self-tuning parameter setting module: used to set iteratively updated self-tuning parameters for each reference signal using a pre-built filtered-x least mean square algorithm;

[0024] A weighting module is configured to assign a weight to each of the reference signals according to the self-adjusting parameters;

[0025] Filtering module: used for filtering the reference signal according to the weight to obtain a filtered signal;

[0026] Noise reduction module: used to use the filtered signal to offset the noise signal and achieve noise reduction.

[0027] In a third aspect, the present invention provides a system including a processor and a storage medium;

[0028] The storage medium is used to store instructions;

[0029] The processor is configured to operate according to the instructions to execute the steps of the method according to any one of the first aspects.

[0030] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the methods described in the first aspect.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] The adaptive pruning multi-reference noise reduction method provided by the present invention introduces a self-adjusting parameter into the reference signal. By adjusting the self-adjusting parameter, the influence of irrelevant signals on the system can be reduced. It can be regarded as adopting a variable step size strategy for each reference signal. The weight of the reference signal can be adjusted according to the correlation between the reference signal and the error signal, realizing the function of adaptive pruning to reduce system overhead and reduce the negative effects of irrelevant signals on the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a flow chart of the adaptive pruning multi-reference denoising method provided by an embodiment of the present invention;

[0034] Figure 2is a schematic diagram of the SWFXLMS algorithm using J reference signals provided by an embodiment of the present invention;

[0035] Figure 3 : This is a schematic diagram of a Sigmoid function image with a value range of [-5, 5] provided by an embodiment of the present invention;

[0036] Figure 4 2 is a schematic diagram of a noise residual image of the FXLMS algorithm provided by an embodiment of the present invention;

[0037] Figure 5 Schematic diagram of a noise residual image of the SWFXLMS algorithm provided by an embodiment of the present invention;

[0038] Figure 6 1 is a schematic diagram of a time domain waveform of a third reference signal with a sudden change provided by an embodiment of the present invention;

[0039] Figure 7 This is a schematic diagram of a time domain waveform of a second reference signal with a sudden change provided by an embodiment of the present invention;

[0040] Figure 8 1 is a schematic diagram of error signal spectra before and after control of the FXLMS algorithm and the SWFXLMS algorithm under a sudden change in the reference signal provided by an embodiment of the present invention;

[0041] Figure 9 Schematic diagram of the self-adjusting parameter iteration process of the SWFXLMS algorithm provided by an embodiment of the present invention;

[0042] Figure 10 1 is a performance diagram of the SWFXLMS algorithm under three reference signal combinations with different amplitudes provided by an embodiment of the present invention;

[0043] Figure 11 3 is a schematic diagram of comparing error signal spectra before and after reference signal pruning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The technical solution of this patent is further described in detail below in conjunction with specific implementation methods.

[0045] The embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention and are not to be construed as limiting the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0046] Example 1:

[0047] Figure 1This is a flow chart of an adaptive pruning multi-reference noise reduction method provided by the first embodiment of the present invention. This flow chart only shows the logical sequence of the method of this embodiment. In other possible embodiments of the present invention, different methods may be used without conflict. Figure 1 The steps shown or described are accomplished in the order shown.

[0048] The adaptive pruning multi-reference noise reduction method provided in this embodiment can be applied to a terminal and can be performed by an adaptive pruning multi-reference noise reduction device. The device can be implemented in software and / or hardware and can be integrated into a terminal, such as any tablet computer or computer device with communication functions. Figure 1 The method of this embodiment specifically includes the following steps:

[0049] Step 1: Using the pre-built filter-x least mean square algorithm, set the iteratively updated self-tuning parameters for each reference signal;

[0050] In this embodiment, a noise control system with a single secondary source and a single error microphone is used. On this basis, it can be expanded to a system with multiple secondary sources and multiple error microphones. Figure 2 FIG. 4 is a schematic diagram of the SWFXLMS algorithm using J reference signals.

[0051] The expression of the filter-x least mean square algorithm is:

[0052]

[0053] Among them, i is the i-th noise source, n is the n-th moment, x i (n) is the reference signal picked up by the sensor at the i-th noise source at the n-th moment, w i (n) is the filter at the nth moment, w i (n+1) is the filter at the n+1th moment, e(n) is the noise residual measured by the error microphone at the nth moment, s(n) is the impulse response of the secondary path at the nth moment, is the estimated value of the impulse response s(n) of the secondary path at the nth moment, λ i (n) is the reference signal x picked up at the i-th noise source at the n-th moment i (n) is the self-adjusting parameter, μ is the step size, and * is the linear convolution operation.

[0054] In this embodiment, the filters all adopt a finite impulse response filter with a transverse structure, and the order of the finite impulse response filter with a transverse structure is M.

[0055] The reference signal x picked up by the sensor at the i-th noise source i The expression for (n) is:

[0056] x i (n) = [x i (n), x i (n-1),…,x i (n-M+1)] T

[0057] Among them, x i (n-1) is the reference signal picked up by the sensor at the i-th noise source at the n-1th moment, x i (n-M+1) is the reference signal picked up by the sensor at the i-th noise source at the n-M+1-th moment, and M is the order of the filter.

[0058] The filter w at the nth moment i The expression for (n) is:

[0059] w i (n)=[w0(n),w1(n),…,w M-1 (n)] T

[0060] Among them, w0(n), w1(n), ..., w M-1 (n) is the 1st, 2nd, ..., Mth coefficient of the filter at the nth moment, and M is the order of the filter.

[0061] The expression of the cancellation signal y(n) generated by the secondary signal through the secondary path is:

[0062]

[0063] Where y(n) is the cancellation signal generated by the secondary signal passing through the secondary path, which is used to cancel the desired signal d(n) to generate the noise residual e(n), and J is the total number of reference signals.

[0064] The noise residual e(n) measured by the error microphone is expressed as:

[0065] e(n)=d(n)-y(n)

[0066] Step 2: Assign weights to each reference signal based on the self-tuning parameters;

[0067] In this embodiment, a self-adjusting parameter between (0, 1) is introduced to assign different weights to each reference signal, so as to reduce the impact of irrelevant signals on the system.

[0068] The expression of the self-tuning parameter is:

[0069]

[0070] Among them, λ i(n) is the reference signal x picked up at the i-th noise source at the n-th moment i (n) self-adjusting parameter, a i (n) is the self-adjusting parameter variable at the nth moment.

[0071] In this embodiment, the self-adjusting parameter adopts the Sigmoid function, exp represents an exponential function with the natural number e as the base, and the function can map the variable to the interval (0, 1), such as Figure 3 As shown, it is the self-adjusting parameter variable a at the nth moment i (n) Schematic diagram of the Sigmoid function image with a value range of [-5, 5].

[0072] However, the iterative update of the self-tuning parameters does not directly modify λ i (n), but through the variable a i (n) Sigmoid function defined to control, λ i (n) and a i (n) there is a positive correlation between them, and by adaptively adjusting a i (n) and then calculate the corresponding λ i (n). The self-adjusting parameter variable a at the nth moment i (n) Adaptive adjustment calculation based on LSM algorithm to obtain:

[0073]

[0074]

[0075] Among them, w1(n), w2(n), ..., w J (n) is the 1st, 2nd, ..., Jth coefficient of the filter at the nth moment, x1(n), x2(n), ..., x J (n) is the reference signal picked up by the sensor at the 1st, 2nd, …, Jth noise source at the nth moment.

[0076] When taking partial derivatives, there are two situations:

[0077]

[0078] Where j is the jth reference signal, λ j (n) is the self-adjusting parameter of the j-th reference signal at the n-th moment.

[0079] According to the above calculation formula, we can get:

[0080]

[0081] Then we get:

[0082]

[0083] The updated variable a i (n) is substituted into the Sigmoid function to obtain the reference signal x picked up at the i-th noise source at the n-th moment i (n) self-tuning parameter λ i (n), and then multiplied by the reference signal on the corresponding path to redistribute the weights to all reference signals.

[0084] The adaptive algorithm with the Sigmoid function weighting has two advantages: First, the reference signal x picked up at the i-th noise source at the n-th moment is i (n) self-tuning parameter λ i (n) is a variable scalar, if we approximate λ i (n) is extracted and combined with the step size μ, the filtered-x least mean square algorithm (SWFXLMS algorithm) can be regarded as an adaptive algorithm with a variable step size for each reference signal. Existing adaptive filtering theory shows that the variable step size algorithm has better convergence speed and stability performance than the fixed step size algorithm.

[0085] Step 3: Filter the reference signal according to the weight to obtain the filtered signal;

[0086] The adaptive pruning multi-reference noise reduction method provided in this embodiment can adjust the weight of each reference signal according to the correlation between the reference signal and the error signal, thereby realizing the adaptive pruning function. Figure 3 It can be seen that the variable a i (n) and λ i (n) there is a positive correlation between i The larger (n) is, the i The larger (n) is, the greater the proportion of the i-th reference signal is. i (n) process, you can The term is approximately regarded as the i-th secondary signal, then:

[0087]

[0088] Among them, y i (n) is the i-th secondary signal.

[0089] Then we get:

[0090] a i (n+1)≈a i (n)+μ a e(n)[y i (n)*s(n)]

[0091] According to the adaptive filtering theory, the variation term depends on e(n)[y i (n)*s(n)], and e(n)[y i (n)*sn is the approximate value of the cross-correlation between the ith secondary signal yin and en, that is, ain is based on y in the iterative process. i The correlation between (n) and e(n) is high or low.

[0092] Step 4: Use the filtered signal to offset the noise signal to achieve noise reduction.

[0093] In order to verify the effectiveness of the adaptive pruning multi-reference denoising method provided in this embodiment, an experiment was conducted to verify it. Figure 4 As shown in, it is a schematic diagram of the noise residual image of the FXLMS algorithm, as Figure 5 Figure 2 shows a schematic diagram of the noise residual image of the SWFXLMS algorithm. The simulation uses 200Hz and 500Hz sinusoidal signals and white noise as three reference signals. The impulse response of the primary path is assumed to be [0.01, 0.25, 0.5, 1, 0.5, 0.25, 0.01], and the impulse response of the secondary path is assumed to be [0.0025, 0.0625, 0.125, 0.25, 0.125, 0.0625, 0.0025]. The desired signal is the convolution of the 200Hz and 500Hz sinusoidal signals with the primary path. Therefore, the white noise reference signal here is an irrelevant reference. The filter order is set to 64. The FXLMS algorithm and SWFXLMS algorithm were used to conduct active control simulation experiments. The step sizes of the two algorithms were set to the maximum values ​​that can ensure the stability of the system: the step size of the FXLMS algorithm was set to 0.0004, the step size of the SWFXLMS algorithm was set to 0.0006, and the self-adjusting parameter variable a at the nth moment was set to 0.0004. i (n) Update step size μ a Set to 0.03. Figure 4 、 Figure 5 It can be seen that the FXLMS algorithm converges slowly; after the active control of the SWFXLMS algorithm, the influence of white noise on the system is reduced. Not only is the convergence speed faster than the FXLMS algorithm, but the noise residual is also significantly smaller.

[0094] To further verify the effectiveness and universality of the adaptive pruning multi-reference noise reduction method provided in this embodiment, vehicle interior noise signals were collected. A dual-channel active noise reduction headrest prototype was fixed to the passenger seat. The collected data included noise signals at both ears of the passenger seat at idle, as well as tachometer and reference microphone signals under normal engine operation. In this case, the primary source of interior noise was engine noise. The binaural test results were similar; in this embodiment, the right ear results are used for discussion.

[0095] In order to obtain a combination of strong correlation signals and weak correlation signals, we spliced ​​the collected engine tachometer signal, reference microphone signal and additional white noise to form four reference signals to simulate the mutation of the reference signal.

[0096] Experiment 1: Verify the performance improvement of the SWFXLMS algorithm.

[0097] The FXLMS algorithm and SWFXLMS algorithm are used to actively control the collected noise, and the step size of each algorithm is set to the maximum value that can ensure the stability of the system. Figure 6 As shown in the figure, it is a schematic diagram of the time domain waveform of the third reference signal with a sudden change, as shown in Figure 7 As shown in FIG, a schematic diagram of the time domain waveform of the second reference signal with a sudden change is shown, where: Figure 6 The first half is the reference signal obtained from the engine tachometer, which has a strong correlation, and the second half switches to the white noise signal with a weak correlation. Figure 7 The first half of the reference signal is white noise, and the second half is the reference signal obtained based on the engine tachometer.

[0098] like Figure 8 The figure shows the error signal spectrum before and after the FXLMS algorithm and the SWFXLMS algorithm are controlled under the sudden change of the reference signal. The dotted line is the signal spectrum before control, the dashed line is the spectrum of the residual noise after the FXLMS algorithm is used, and the solid line is the spectrum of the residual noise after the SWFXLMS algorithm is used. Figure 8 It can be seen that at the main harmonic frequencies, both algorithms can achieve effective noise reduction effects, and the noise reduction effect of the SWFXLMS algorithm is about 3dB better than that of the FXLMS algorithm, and its noise reduction effect is better.

[0099] The SWFXLMS algorithm adjusts parameters according to the correlation between different reference signals and noise, and reduces the impact on the system by reducing the self-adjusting parameters of the irrelevant signal part. Figure 9 As shown in the figure, it is a schematic diagram of the self-adjusting parameter iteration process of the SWFXLMS algorithm. Figure 9 It can be seen that when the third reference signal suddenly changes from a relevant signal to an irrelevant signal, λ i (n) is gradually iteratively reduced. When the second reference signal suddenly changes from an irrelevant signal to a relevant signal, λ i (n) has changed significantly and its weight has increased, so compared with the FXLMS algorithm, the SWFXLMS algorithm can highlight the reference signals with strong correlation by adjusting the weights of the reference signals.

[0100] Experiment 2: Investigate the extent to which the performance of the SWFXLMS algorithm is affected by the amplitude of the reference signal, that is, whether the noise reduction effect of the SWFXLMS algorithm is stable when the amplitude of the reference signal changes dynamically.

[0101] The amplitude of the irrelevant signal in the first experiment was reset, and the collected signal was combined with irrelevant signals of smaller amplitude, similar amplitude, and larger amplitude respectively. The three combinations with different signal amplitude differences were actively controlled using the SWFXLMS algorithm, such as Figure 10 The figure shows the performance of the SWFXLMS algorithm under three combinations of reference signals with different amplitudes. Because the SWFXLMS algorithm is sensitive to irrelevant signals, even if the amplitude differences between the reference signals are large, the self-adjusting parameters on the irrelevant reference signals can be gradually iterated to the optimal solution, and finally the irrelevant signals are given smaller weights to reduce the impact on the system. The quality of the final noise reduction effect is mainly related to the high-correlation signal. Therefore, Figure 10 The frequency spectrum of the error signal is basically the same.

[0102] Test 3: Verification basis λ i (n) The stability of the SWFXLMS algorithm after the dynamic pruning of the reference signal, that is, if λ i When (n) is less than the set threshold, the reference signal does not participate in the generation of the secondary signal. Here, the threshold is set to 0.15.

[0103] like Figure 11 As shown in the figure, it is a schematic diagram of the comparison of the error signal spectrum before and after the reference signal is pruned. Figure 11 It can be seen that after pruning, the noise reduction performance of the system does not deteriorate significantly, and the computational overhead of the system can be reduced after pruning.

[0104] The adaptive pruning multi-reference noise reduction method provided in this embodiment uses the SWFXLMS algorithm to weight different reference signals using a Sigmoid function. The algorithm reduces the impact of irrelevant signals on the system by introducing self-adjusting parameters in the reference signal path. Simulation experimental results show that even if there are large amplitude differences between the reference signals, the SWFXLMS algorithm can still ensure a stable noise reduction effect and effectively remove the interference of irrelevant noise. Moreover, the self-adjusting parameters introduced by the algorithm are simple in form and require less computation, making them easy to implement in hardware and can be well applied to adaptive noise control systems.

[0105] Example 2:

[0106] This embodiment provides an adaptive pruning multi-reference noise reduction device, including:

[0107] Self-tuning parameter setting module: used to set iteratively updated self-tuning parameters for each reference signal using a pre-built filtered-x least mean square algorithm;

[0108] Weighting module: used to assign weights to each reference signal based on self-adjusting parameters;

[0109] Filtering module: used to filter the reference signal according to the weight to obtain the filtered signal;

[0110] Noise reduction module: used to use the filtered signal to offset the noise signal and achieve noise reduction.

[0111] The adaptive pruning multi-reference denoising device provided in the embodiment of the present invention can execute the adaptive pruning multi-reference denoising method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0112] Example 3:

[0113] This embodiment provides a system, including a processor and a storage medium;

[0114] The storage medium is used to store instructions;

[0115] The processor is configured to operate according to the instructions to execute the steps of the method in the first embodiment.

[0116] Example 4:

[0117] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the method in the first embodiment are implemented.

[0118] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0119] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0120] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0121] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0122] The above are only preferred embodiments of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An adaptive pruning multi-reference denoising method, characterized in that: include: Using the pre-built filter-x least mean square algorithm, the self-tuning parameters for each reference signal are set for iterative update; Assign weights to each reference signal based on self-tuning parameters; Filter the reference signal according to the weight to obtain a filtered signal; Use the filtered signal to offset the noise signal to achieve noise reduction; The expression of the self-tuning parameter is: ； in, For the The moment The reference signal picked up at the noise source The self-tuning parameters, For the The self-adjusting parameter variables at each moment; The iterative update formula of the self-adjusting parameter variable is: ； in, For the The self-adjusting parameter variable at each moment, For the The self-adjusting parameter variable at each moment, For the The self-adjusting parameter variable at each moment The step size during update, For the error microphone The noise residual measured at each moment, For the The filter at each moment, For the sensor The moment The reference signal picked up at a noise source, For the The moment The reference signal picked up at the noise source The self-tuning parameters, For the The impulse response of the secondary path at time instant is, It is a linear convolution operation.

2. The adaptive pruning multi-reference denoising method according to claim 1, characterized in that: The expression of the filter-x least mean square algorithm is: ； in, For the Noise sources, For the A moment, For the sensor The moment The reference signal picked up at a noise source, For the The filter at each moment, For the The filter at each moment, For the error microphone The noise residual measured at each moment, For the The impulse response of the secondary path at time instant is, For the The impulse response of the secondary path at time The estimated value of For the The moment The reference signal picked up at the noise source The self-tuning parameters, is the step length, It is a linear convolution operation.

3. The adaptive pruning multi-reference denoising method according to any one of claims 1 or 2, characterized in that: Each of the filters adopts a finite impulse response filter with a transverse structure, and the order of the finite impulse response filter of the transverse structure is .

4. An adaptive pruning multi-reference noise reduction device, characterized in that: include: Self-tuning parameter setting module: used to set iteratively updated self-tuning parameters for each reference signal using a pre-built filtered-x least mean square algorithm; A weighting module is configured to assign a weight to each of the reference signals according to the self-adjusting parameters; Filtering module: used for filtering the reference signal according to the weight to obtain a filtered signal; Noise reduction module: used to use the filtered signal to offset the noise signal and achieve noise reduction; The expression of the self-tuning parameter is: ； in, For the The moment The reference signal picked up at the noise source The self-tuning parameters, For the The self-adjusting parameter variables at each moment; The iterative update formula of the self-adjusting parameter variable is: ； in, For the The self-adjusting parameter variable at each moment, For the The self-adjusting parameter variable at each moment, For the The self-adjusting parameter variable at each moment The step size during update, For the error microphone The noise residual measured at each moment, For the The filter at each moment, For the sensor The moment The reference signal picked up at a noise source, For the The moment The reference signal picked up at the noise source The self-tuning parameters, For the The impulse response of the secondary path at time instant is, It is a linear convolution operation.

5. A system, characterized in that: including processors and storage media; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images