Filter parameter optimization method and system, noise reduction method and Bluetooth earphone

Abstract

The invention discloses a filter parameter optimization method. The method is characterized in that a first feed-forward microphone, a second feed-forward microphone and a virtual microphone are respectively arranged; the first feed-forward microphone acquires a first signal, the second feed-forward microphone acquires a second signal, and the virtual microphone acquires an expected signal; adding and mixing the first signal and the second signal after passing through the first feed-forward controller and the second feed-forward controller to obtain an adding and mixing signal; carrying out difference mixing on the addition mixed signal and the expected signal through a secondary path to obtain a residual signal; and respectively adjusting filter parameters of the first feed-forward controller and the second feed-forward controller through the residual information. Parameters of the active noise reduction filter can be adjusted according to external noise conditions and user requirements, so that good noise reduction effect and audio quality are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of active noise reduction of Bluetooth headsets, and in particular to a filter parameter optimization method, a system, a noise reduction method and a Bluetooth headset. Background Art

[0002] Bluetooth headsets have become an indispensable part of people's daily lives, and active noise cancellation (ANC) is one of the important functions of Bluetooth headsets. When using Bluetooth headsets, the active noise cancellation function can help users reduce the interference of external noise, improve the quality of audio signals, and thus provide a better listening experience. In order to obtain better active noise cancellation effects, high-end Bluetooth headsets are usually equipped with multiple microphones and multiple speakers to achieve a quieter noise cancellation experience. Multi-channel hybrid ANC (MCHANC) headsets include feedforward (FF) and feedback (FB) structures, and their configuration is M*N*P, where M is the number of feedforward microphones, N is the number of speakers, and P is the number of feedback microphones. At present, there are few active noise cancellation parameter debugging schemes for multi-channel hybrid Bluetooth headsets. Most of the existing schemes are (Filter-x Least Mean Square, FxLMS) algorithms for multi-channel feedforward Bluetooth headsets, but lack parameter debugging research for multi-channel hybrid. At the same time, the existing target noise reduction positions are all at the location where the feedback microphone is located, while the actual noise reduction position is at the human ear, so there is more or less noise reduction error. Summary of the invention

[0003] Based on the above situation, the present invention proposes to test the transfer function data of ANC headphones through grouping, including obtaining primary path test data and secondary path test data; obtaining FB filter parameters based on the transfer function data of the ANC headphones and a swarm intelligent optimization algorithm; and obtaining FF filter parameters based on the transfer function data, FB filter parameters, and FxLMS algorithm.

[0004] The present invention provides a filter parameter optimization method, by respectively setting a first feedforward microphone, a second feedforward microphone, a feedback microphone and a virtual microphone; the first feedforward microphone obtains a first signal, the second feedforward microphone obtains a second signal, the feedback microphone obtains a first expected signal and the virtual microphone obtains a second expected signal;

[0005] The first signal passes through a first feedforward controller and the second signal passes through a second feedforward controller and then performs addition and mixing to obtain an addition mixed signal; the addition mixed signal passes through a first secondary path and is differentially mixed with a first desired signal to obtain a first residual signal; the first residual signal passes through a feedback controller to obtain a residual compensation signal, and the residual compensation signal is added and mixed with the addition mixed signal to form an addition residual compensation signal; and the second desired signal passes through a second secondary path and is differentially mixed with the addition residual compensation signal to obtain a second residual signal.

[0006] Furthermore, the transfer function of the first secondary path is obtained by dividing the feedback microphone data and the speaker data after Fourier transformation, and the transfer function of the second secondary path is obtained by dividing the virtual microphone data and the speaker data after Fourier transformation.

[0007] Furthermore, the first feedforward microphone and the second feedforward microphone are in a left-right or up-down configuration and are distributed along a straight line.

[0008] Furthermore, the sound pickup holes of the first feed-forward microphone and the second feed-forward microphone are opened in opposite directions.

[0009] The present invention provides an active noise reduction method for headphones, which comprises the following steps: a first feedforward microphone, a second feedforward microphone, a feedback microphone and a virtual microphone are respectively arranged; the first feedforward microphone obtains a first signal, the second feedforward microphone obtains a second signal, the feedback microphone obtains a first expected signal and the virtual microphone obtains a second expected signal; the first signal passes through a first feedforward controller and the second signal passes through a second feedforward controller and then is added and mixed to obtain an added mixed signal; the added mixed signal passes through a first secondary path and is differentially mixed with the first expected signal to obtain a first residual signal; the first residual signal passes through a feedback controller to obtain a residual compensation signal, the residual compensation signal is added and mixed with the added mixed signal to form an added residual compensation signal; the added residual compensation signal is then used to pass through a second secondary path and is differentially mixed with the second expected signal to perform active noise reduction.

[0010] At the same time, the present invention also provides a filter parameter optimization system, including: a signal acquisition module, a signal processing module and an output module, and each module is signal-connected.

[0011] The signal acquisition module is used to first set the first feedforward microphone, the second feedforward microphone, the feedback microphone and the virtual microphone respectively; wherein the first feedforward microphone acquires the first signal, the second feedforward microphone acquires the second signal, the feedback microphone acquires the first expected signal and the virtual microphone acquires the second expected signal; the signal processing module is used to add and mix the first signal after passing through the first feedforward controller and the second signal after passing through the second feedforward controller to obtain the added mixed signal; the added mixed signal is mixed with the first expected signal after passing through the first secondary path to obtain the first residual signal; the first residual signal is then passed through the feedback controller to obtain the residual compensation signal, the residual compensation signal is added and mixed with the added mixed signal to form the added residual compensation signal; the added residual compensation signal is mixed with the second expected signal after passing through the second secondary path to obtain the second residual signal; the parameter adjustment module is used to adjust the filter parameters of the feedback controller according to the first residual signal; and the filter parameters of the first feedforward controller and the second feedforward controller are adjusted according to the second residual signal.

[0012] Some technical effects of the present disclosure are: the parameters of the active noise reduction filter can be adjusted according to the external noise situation and user needs, so as to achieve the best noise reduction effect and audio quality. The method proposed by the present invention will provide technical support for the design and production of ANC Bluetooth headsets, improve product competitiveness and user satisfaction. At the same time, the headset can be actively noise-reduced to reduce the interference of external noise on the audio playback effect in the headset. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] To better understand the technical solution of the present disclosure, you may refer to the following drawings for auxiliary explanation of the prior art or embodiments. These drawings will selectively display the products or methods involved in the prior art or some embodiments of the present disclosure. The basic information of these drawings is as follows:

[0014] Figure 1 It is a flow chart of an embodiment of a filter parameter optimization method of the present invention;

[0015] Figure 2 It is a schematic diagram of embodiment 1 of the noise reduction data acquisition path of the present invention.

[0016] Figure 3 It is a schematic diagram of simulation results of the coherence of the dual feedforward paths of the present invention.

[0017] Figure 4 It is a schematic diagram of the microphone layout position of the multi-feedforward controller of the present invention.

[0018] Figure 5 It is a schematic diagram of embodiment 2 of the noise reduction data acquisition path of the present invention.

[0019] Figure 6It is a structural schematic diagram of an embodiment of a filter parameter optimization system of the present invention. DETAILED DESCRIPTION

[0020] The technical means or technical effects involved in the present disclosure will be further described below. Obviously, the embodiments provided are only some of the embodiments of the present disclosure, not all of them. Based on the embodiments in the present disclosure and the explicit or implicit descriptions of the figures and texts, all other embodiments that can be obtained by those skilled in the art without making creative efforts will be within the scope of protection of the present disclosure.

[0021] As one of the embodiments, the transfer function data of the ANC headset is tested in groups, including obtaining primary path test data and secondary path test data; obtaining FB filter parameters based on the transfer function data of the ANC headset and the swarm intelligence optimization algorithm; and obtaining FF filter parameters based on the transfer function data, FB filter parameters, and FxLMS algorithm. For ease of representation, the relevant steps of the present invention are described in the (2,1,1) configuration of M=2, N=1, and P=1. In particular, the frequency domain form of the signal X is represented as X(z), and its corresponding time domain form is represented as X(n).

[0022] like Figure 1 As shown, the method in this embodiment includes the steps of:

[0023] S1, respectively setting a first feedforward microphone, a second feedforward microphone, a feedback microphone and a virtual microphone; the first feedforward microphone obtains a first signal, the second feedforward microphone obtains a second signal, the feedback microphone obtains a first desired signal and the virtual microphone obtains a second desired signal;

[0024] Lead out the positive and negative leads of the two feedforward microphones and one feedback microphone of the prototype (headphones), the data line, and the positive and negative leads of the speaker; and place a virtual microphone in front of the eardrum of the artificial ear. The models of the feedforward microphone, feedback microphone, and virtual microphone should be kept the same as much as possible. In the anechoic room, in an environment where pink noise is played outside, download the data of the feedforward microphone, feedback microphone, and virtual microphone at the same time. The data downloaded by the first feedforward microphone is the first signal x1(n), the data downloaded by the second feedforward microphone is the second signal x2(n), the data downloaded by the feedback microphone is the time domain form Derp(n) of the first expected signal Derp(z), and the data downloaded by the virtual microphone is the time domain form Ddrp(n) of the second expected signal Ddrp(z).

[0025] Derp(n) is the time domain signal received by the feedback microphone. It is also directly divided into the primary path test data. It is mainly used for the following Figure 1The signal of the target noise reduction point during the operation of the mixed-feedback headset ANC cannot be directly obtained, but only the signal of the feedback microphone can be directly obtained. Therefore, the signal of the feedback microphone needs to be used to map the signal of the virtual microphone. Therefore, the time-frequency form of the signal received by the feedback microphone is used in this embodiment, which is mainly used for the modeling part later.

[0026] like Figure 2 As shown, this embodiment divides the data acquisition path into a primary path and a secondary path. The process of acquiring the primary path test data and the secondary path test data: the primary path refers to the path from the sound source to the target noise reduction point, and the secondary path refers to the path from the speaker to the target noise reduction point, which is divided into a first secondary path and a second secondary path, wherein the first secondary path is the path from the speaker to the feedback microphone, and the second secondary path is the path from the speaker to the virtual microphone.

[0027] The data required for the primary path test data include: when pink noise is played outside, the first signal x1(n) (time domain signal) collected by the first feedforward microphone 1, the second signal x2(n) (time domain signal) collected by the second feedforward microphone 2, the data downloaded by the feedback microphone is the time domain form Derp(n) of the first expected signal Derp(z), and the data downloaded by the virtual microphone is the time domain form Ddrp(n) of the second expected signal Ddrp(z). In the anechoic room, under the pink noise environment played by the speaker in the earphone, the data of the speaker, feedback microphone, and virtual microphone are downloaded at the same time. Speaker playback Figure 2 The data after the sum of the signals, the transfer function of the first secondary path SPerp(z) can be obtained by dividing the feedback microphone and speaker data after Fourier transform, and the transfer function of the second secondary path SPdrp(z) can be obtained by dividing the virtual microphone and speaker data after Fourier transform, which is a fixed path for obtaining corresponding data during modeling. The summed mixed signal after passing through the first secondary path is the summed mixed signal multiplied by the transfer function of the first secondary path; the transfer function of the second secondary path is obtained by dividing the virtual microphone data and the speaker data after Fourier transform; the summed residual compensation signal after passing through the second secondary path is the summed residual compensation signal multiplied by the transfer function of the second secondary path. This is a more conventional processing method in the field of acoustics, and will not be described in detail in this embodiment.

[0028] The patent of this invention only needs the frequency domain form of the secondary path test data, so only the first secondary path SPerp(z) and the second secondary path SPdrp(z) shown in the figure are given. Now the detailed calculation process of the above two secondary paths is given. The signal of pink noise played inside the earphone is spk(n), at this time the data received by the feedback microphone is Derp1(n), and the data received by the virtual microphone is Ddrp1(n) (different from the data received by the primary path), and then using these two time domain signals, the first secondary path SPerp(z)=fft(Derp1(n) / spk(n)) and the second secondary path SPdrp(z)=fft(Ddrp1(n) / spk(n)) can be obtained by discrete Fourier transform calculation.

[0029] S2, the first signal passes through the first feedforward controller and the second signal passes through the second feedforward controller and then is added and mixed to obtain a summed mixed signal; the summed mixed signal passes through the first secondary path and is differentially mixed with the first desired signal to obtain a first residual signal; the first residual signal passes through the feedback controller to obtain a residual compensation signal, the residual compensation signal is added and mixed with the summed mixed signal to form a summed residual compensation signal; the summed residual compensation signal is then used to pass through the second secondary path and is differentially mixed with the second desired signal to obtain a second residual signal.

[0030] It includes two parts, a feedforward path and a feedback path, wherein the composition of the feedforward path can be expressed as follows: the input signal x1(n) and the signal x2(n) are added and mixed respectively through the feedforward controller FF1(z) and the feedforward controller FF2(z), and then the first one is played through the secondary path SPerp(z) and is differentially mixed with the first desired signal Derp(z) at the feedback microphone, and the second one is played through the second secondary path SPdrp(z) and is differentially mixed with the second desired signal Ddrp(z) at the virtual microphone; wherein the composition of the feedback path can be expressed as follows: the first residual signal Eerp(z) passes through the feedback controller FB(z), and then the first one is played through the first secondary path SPerp(z) and is superimposed with the second desired signal Derp(z) at the feedback microphone to form a closed loop, and the second one is played through the second secondary path SPdrp(z) and is superimposed with the first desired signal Ddrp(z) at the virtual microphone. Considering that the eardrum is the actual noise reduction point and the feedback microphone (physical microphone) is the point where the headphone processes the residual signal, the two do not overlap, so the virtual microphone and the feedback microphone (physical microphone) are modeled separately.

[0031] The relationship between the channels can be expressed as:

[0032] Eerp(z)=Derp(z)-(x1(z)*FF1(z)+x2(z)*FF2(z))*SPerp(z)-Eerp(z)*FB(z)*SPerp(z)

[0033] Edrp(z)=Ddrp(z)-(x1(2)*FF1(z)+x2(z)FF2(z))*SPdrp(z)-Eerp(z)*FB(z)*SPdrp(z)

[0034] Ddrp(z)=Derp(z)*PPe2d(z)

[0035] Where PPe2d(z) represents the transfer function from the feedback microphone point to the virtual microphone point.

[0036] Arrange the above formula:

[0037] Edrp(z)=Ddrp(z)-PART1(z)-PART2(z)

[0038] in

[0039]

[0040] Among them, PART1(z) is the anti-phase noise signal played by the speaker through SPdrp(z) when only the feedback path is working, and PART1(z)+PART2(z) is the anti-phase noise signal played by the speaker through SPdrp(z) when the mixed feedback path is working. The goal of active noise reduction is to make the amplitude of PART1(z)+PART2(z) as close to Ddrp(z) as possible and the phase as opposite to Ddrp(z) as possible.

[0041] According to the above formula, when the mixed feed noise reduction works, the noise reduction effect is jointly affected by the feedback controller FB(z), the feedforward controllers FF1(z), and FF2(z), and the noise reduction effect of the feedforward control is affected by the feedback control. Therefore, in the present invention, it is necessary to first determine the parameterized filter corresponding to FB(z) and then determine the parameterized filter corresponding to the feedforward controllers FF1(z) and FF2(z).

[0042] S3. Adjust the filter parameters of the feedback controller according to the first residual signal; and adjust the filter parameters of the first feedforward controller and the second feedforward controller respectively according to the second residual signal.

[0043] The methods disclosed in patents CN118430500A and CN116743113A can be applied to the process of obtaining the FB filter in the present invention. Specifically, it is assumed that the feedback controller consists of K second-order IIRs, specifically expressed as [Gfb, gfb1, gfb1, Fcfb1, ..., gfbk, qfbk, Fcfbk, ..., gfbk, qfbK, FcfbK], where Gfb is the link gain, gfbk, gfbk, Fcfbk are the parameters of the kth second-order IIR, corresponding to the gain, quality factor, and center frequency of the IIR, respectively. The open-loop transfer function FB(z)*SPerp(z) can be constructed, is the cost function of the target noise reduction amount, and the parameters of K second-order IIRs are iterated through a swarm intelligence optimization algorithm. In this embodiment, the acquisition of the feedback part filter in the mixed feed link is not described in detail.

[0044] Optionally, in order to obtain a better noise reduction effect for the mixed-feed link, this embodiment first provides a method for placing the reference microphone, in which multi-path coherence is used to evaluate the advantages and disadvantages of the reference microphone placement. Specifically, since the present invention involves multi-path mixed-feed active noise reduction, the coherence path evaluated by the present invention is the coherence between the input signal x1(n), the input signal x2(n) and the desired signal Drp(n) of the virtual microphone point. The coherence calculation formula under the dual-input single-output MISO model is:

[0045]

[0046] Where P ab represents the calculation of the power spectral density between signal a and signal b, (·) * represents the conjugate, (·) -1 In particular, when there is only one input, the above formula degenerates into

[0047]

[0048] in The signal x1(n) corresponds to red, x2(n) corresponds to green, and Drp(n) corresponds to blue, and inputs C Xy The simulation results of MISO can be obtained by using the calculation formula, such as Figure 3 shown.

[0049] Figure 3 The results show that the coherence of the dual feedforward path is improved compared to the upper limit of the coherence of the single feedforward path, which means that compared with a single FF, the application of dual FF can improve the limit of feedforward noise reduction.

[0050] As a preferred implementation example, a diagram of the positions of multiple feedforward microphones on the headphones is also given. If it is a double FF, it is usually in a left-right or up-down configuration, distributed along the diameter. If it is a triple FF, it is three points with an interval of 120°. In order to ensure that the coherence between the microphones meets the requirements, the sound pickup holes of the general microphones need to be opened in opposite directions, such as Figure 4 shown.

[0051] The exemplary multi-channel feedforward microphone involved in this embodiment is a dual FF. The positions of the two reference microphones can be determined by the method shown in this embodiment. The method for obtaining the parameters of the FB filter has been given in step 1. Now the method for obtaining the feedforward filter parameters FF1(z) and FF2(z) is given: Considering that Edrp(z) = Ddrp(z)-PART1(z)-PART2(z), when the FB coefficient has been determined, Ddrp(z) and PART1(z) are both fixed values, and PART1(z)

[0052] is also a fixed value, so we can consider using the FXNLMS algorithm on the time domain signal of Edrp(z) to obtain the target frequency response of the feedforward filter. Rewrite Edrp(z) into the following time domain form:

[0053] Edrp(n)=Ddrp(n)-PART1(n)-PART2(n)

[0054] Where Ddrp(n) is the time domain form of Ddrp(z), which is the signal received by the virtual microphone when the active noise reduction is turned off, that is, the expected signal, and Edrp(n) is the residual signal received by the virtual microphone when the active noise reduction is working. PART1(n) can be expressed as

[0055]

[0056] Where filter() is the filter function in matlab / python, and ifft() is the inverse Fourier transform function. Derp(n) is the signal received by the feedback microphone when the active noise reduction is turned off. PART2(n) can be expressed as

[0057] PART2(n)=filterx1(n)+filterx2(n)

[0058] in,

[0059]

[0060] The reason for the above formula is that if the FxLMS algorithm is used, the target point must be converted into the corresponding time domain form. Figure 2 Mixed feed dual FF feedforward convergence target:

[0061] min{J=E[Edrp(n) 2 ]}

[0062] Assume that the FIR corresponding to the feedforward controllers FF1(z) and FF2(z) are represented as w1 and w2 respectively; the time domain input sequences of x1 and x2 after pre-filtering are represented as filterx1 and filterx2 respectively, where

[0063] w1=[w1(1),...,w1(n),...,w1(N)]

[0064] w2=[w2(1),...,w2(n),...,w2(N)]

[0065] filterx1=[filterx1(1),...,filterx1(n),...,filterx1(N)]

[0066] filterx2=[filterx2(1),...,filterx2(n),...,filterx2(N)]

[0067] w1, w2, filterx1, and finlterx2 represent time domain input sequences of length N.

[0068] If the convergence goal is met, then,

[0069]

[0070] The corresponding FxLMS solution is:

[0071] w1(n+1)=w1(n)+μ1e(n)filterx1(n)

[0072] w2(n+1)=w2(n)+μ2e(n)filterx1(n)

[0073] Wherein, μ1 and μ2 are both constants, and the present invention provides example values ​​of μ1 and μ2: μ1=20, μ2=20.

[0074] e(n) is the time domain sequence of the residual at the nth moment. Specifically,

[0075] e(n)=Ddrp(n)-PART1(n)-(fllterx1(n)*w1(n)+filterx2(n)*w2(n))

[0076] Optionally, the present invention provides a condition satisfying min{J=E[Edrp(n) 2 ]}The corresponding FxNLMS solution is:

[0077] μ1=alpha / filterx1(n) T filterx1(n)

[0078] μ2=alpha / filterx2(n) T filterx2(n)

[0079] w1(n+1)=w1(n)+μ1e(n)filterx1(n)

[0080] w2(n+1)=w2(n)+μ2e(n)filterx1(n)

[0081] Wherein, alpha is a constant and 0<alpha<1. The present invention provides an example value of alpha: alpha=0.0128.

[0082] So far, the present invention has obtained the feedforward controllers FF1(z) and FF2(z), and the corresponding filter parameters are respectively expressed as w1 and w2. Assuming that the parameterized filter corresponding to the feedforward controller FF1(z) can be expressed as [Gff1, gff11, qff11, Fcff11, ..., gff1k, qff1k, Fcff1k, ..., gff1K, qff1K, Fcff1K], and the parameterized filter corresponding to the feedforward controller FF2(z) can be expressed as [Gff2, gff21, qff21, Fcff21, ..., gff2k, qff1k, Fcff2k, ..., gff2K, qff2K, Fcff2K], the process of converting from w1 and w2 to the parameterized filter can be expressed as the following steps:

[0083] Calculate the frequency responses hw1(z), hw2(z) corresponding to w1 and w2. The process of obtaining the frequency responses can be calculated by the matlab / python freqz function. Given the initial value of the parameterized filter, it is assumed in the present invention that the parameterized filters corresponding to the feedforward controller FF1(z) and the feedforward controller FF2(z) are [0, 0, 1, 1000, ..., 0, 1, 1000, ..., 0, 1, 1000], and all K filters are peak filters. Obtain the frequency responses hw1(z), hw2(z) corresponding to the filters.

[0084] The filter is iterated by swarm intelligence optimization algorithms, such as genetic algorithm, particle swarm optimization algorithm, simulated annealing algorithm, etc. The iterative objectives are: min((hw1(z)-hwiir1(z)) 2 *Weight1(z)), min((hw2(z)-hwiir2(z)) 2*Weight2(z)), the present invention provides an example weight setting method: Weight2(z)=Weight1(z)=1.

[0085] So far, this embodiment has obtained the filter parameters corresponding to the feedforward controllers FF1(z) and FF2(z) and the parameters corresponding to the feedback controller F(z).

[0086] As another example, Figure 5 As shown, a filter parameter optimization method includes the following steps: respectively setting a first feedforward microphone, a second feedforward microphone and a virtual microphone; the first feedforward microphone obtains a first signal, the second feedforward microphone obtains a second signal, and the virtual microphone obtains an expected signal; the first signal passes through a first feedforward controller and the second signal passes through a second feedforward controller and then is added and mixed to obtain an added mixed signal; the added mixed signal is differentially mixed with the expected signal through a secondary path to obtain a residual signal; and the filter parameters of the first feedforward controller and the second feedforward controller are adjusted respectively through the residual signal. Compared with the previous embodiment, the first secondary path is reduced, and the feedback path corresponding to the feedback microphone is reduced. One residual compensation signal is reduced, but it can also achieve a better parameter optimization effect, while reducing the corresponding hardware cost and algorithm calculation cost. At this time, the filter parameter update of the first feedforward controller and the second feedforward controller can be achieved by replacing the following formula:

[0087] Edrp(n)=Ddrp(n)-PART1(n)-PART2(n)

[0088] FB(z) in is set to 0. Figure 5 The links shown are considered Figure 2 The state in which the feedback controller in the link is not working.

[0089] As another example, a method for active noise reduction of headphones includes:

[0090] A first feedforward microphone, a second feedforward microphone, a feedback microphone and a virtual microphone are respectively set; the first feedforward microphone obtains a first signal, the second feedforward microphone obtains a second signal, the feedback microphone obtains a first expected signal and the virtual microphone obtains a second expected signal; the first signal passes through a first feedforward controller and the second signal passes through a second feedforward controller and then is added and mixed to obtain an added mixed signal; the added mixed signal passes through a first secondary path and is differentially mixed with the first expected signal to obtain a first residual signal; the first residual signal passes through a feedback controller to obtain a residual compensation signal, the residual compensation signal is added and mixed with the added mixed signal to form an added residual compensation signal; the second expected signal passes through a second secondary path and is differentially mixed with the added residual compensation signal to perform noise reduction. Since the second residual signal is essentially an active noise reduction for external sounds, the process of obtaining the second residual signal is the process of active noise reduction.

[0091] Preferably, before use, the filter parameter optimization method of embodiment 1 is used to optimize the parameters of the corresponding filter.

[0092] As another implementation example, a method for active noise reduction of headphones includes: respectively setting a first feedforward microphone, a second feedforward microphone and a virtual microphone; the first feedforward microphone obtains a first signal, the second feedforward microphone obtains a second signal, and the virtual microphone obtains an expected signal; the first signal passes through a first feedforward controller and the second signal passes through a second feedforward controller and then is added and mixed to obtain a summed mixed signal; the summed mixed signal is differentially mixed with the expected signal through a secondary path to perform active noise reduction.

[0093] The secondary path here is equivalent to the second secondary path in the previous embodiment, and the second secondary path is the path from the speaker to the virtual microphone. The transmission function of the second secondary path is obtained by dividing the virtual microphone data and the speaker data after Fourier transformation; the summed mixed signal is mixed with the desired signal through the secondary path to obtain a signal to the human ear, which is the signal after active noise reduction.

[0094] Preferably, before use, the filter parameter optimization method of embodiment 2 is used to optimize the parameters of the corresponding filter.

[0095] like Figure 6 As shown, this embodiment also includes a filter parameter optimization system, including: a signal acquisition module, a signal processing module and a parameter adjustment module, and each module is signal-connected.

[0096] The signal acquisition module is used to respectively set a first feedforward microphone, a second feedforward microphone, a feedback microphone and a virtual microphone; wherein the first feedforward microphone acquires a first signal, the second feedforward microphone acquires a second signal, the feedback microphone acquires a first desired signal and the virtual microphone acquires a second desired signal;

[0097] The signal processing module is used to add and mix the first signal after passing through the first feedforward controller and the second signal after passing through the second feedforward controller to obtain an added mixed signal; then the added mixed signal is differentially mixed with the first desired signal after passing through the first secondary path to obtain a first residual signal; the first residual signal is then passed through the feedback controller to obtain a residual compensation signal, the residual compensation signal is added and mixed with the added mixed signal to form an added residual compensation signal; the added residual compensation signal is differentially mixed with the second desired signal after passing through the second secondary path to obtain a second residual signal;

[0098] The parameter adjustment module is used to adjust the filter parameters of the feedback controller according to the first residual signal; and to adjust the filter parameters of the first feedforward controller and the second feedforward controller respectively through the second residual signal.

[0099] An embodiment of a human voice detection and speech recognition system based on a low-resource scenario is based on a method embodiment, which will not be described repeatedly here.

[0100] As an application of this embodiment, this embodiment also provides a Bluetooth headset, including: a Bluetooth headset body, the headset body adopts the filter parameter optimization method as described above. As a simple replacement, this Bluetooth headset body can also be used in other products, such as the headset body of a wired headset, etc., which are all simple replacements of this embodiment.

[0101] Within the scope of the knowledge and ability level of those skilled in the art, the various embodiments or technical features mentioned herein may be combined with each other as other optional embodiments without conflict. These limited number of optional embodiments, which are not listed one by one and are formed by combining a limited number of technical features, still fall within the technical scope disclosed in the present disclosure and can be understood or inferred by those skilled in the art in combination with the drawings and the above text.

[0102] It is emphasized again that the embodiments listed above are typical and preferred embodiments of the present disclosure, which are only used to explain and interpret the technical solutions of the present disclosure in detail to facilitate the understanding of the readers, and are not intended to limit the protection scope or application of the present disclosure. Any technical solutions obtained by any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A filter parameter optimization method, characterized in that: S1, respectively setting a first feedforward microphone, a second feedforward microphone, a feedback microphone and a virtual microphone; the first feedforward microphone obtains a first signal, the second feedforward microphone obtains a second signal, the feedback microphone obtains a first desired signal and the virtual microphone obtains a second desired signal; S2, the first signal passes through a first feedforward controller and the second signal passes through a second feedforward controller and then is added and mixed to obtain an added mixed signal; the added mixed signal passes through a first secondary path and is differentially mixed with a first desired signal to obtain a first residual signal; the first residual signal passes through a feedback controller to obtain a residual compensation signal, the residual compensation signal is added and mixed with the added mixed signal to form an added residual compensation signal; the added residual compensation signal passes through a second secondary path and is differentially mixed with the second desired signal to obtain a second residual signal; S3. Adjust the filter parameters of the feedback controller according to the first residual signal; and adjust the filter parameters of the first feedforward controller and the second feedforward controller respectively according to the second residual signal.

2. The filter parameter optimization method according to claim 1, characterized in that: The first feedforward microphone and the second feedforward microphone are in a left-right or up-down configuration and are distributed along a straight line.

3. The filter parameter optimization method according to claim 1 or 2, characterized in that: The sound pickup holes of the first feed-forward microphone and the second feed-forward microphone are opened in opposite directions.

4. A filter parameter optimization method, characterized in that: A first feedforward microphone, a second feedforward microphone and a virtual microphone are respectively set; the first feedforward microphone obtains a first signal, the second feedforward microphone obtains a second signal, and the virtual microphone obtains an expected signal; the first signal passes through a first feedforward controller and the second signal passes through a second feedforward controller and then is added and mixed to obtain an added mixed signal; the added mixed signal is differentially mixed with the expected signal through a secondary path to obtain a residual signal; and the filter parameters of the first feedforward controller and the second feedforward controller are respectively adjusted according to the residual information.

5. A method for active noise reduction of headphones, characterized by: A first feedforward microphone, a second feedforward microphone, a feedback microphone and a virtual microphone are respectively set; the first feedforward microphone obtains a first signal, the second feedforward microphone obtains a second signal, the feedback microphone obtains a first expected signal and the virtual microphone obtains a second expected signal; the first signal passes through a first feedforward controller and the second signal passes through a second feedforward controller and are added and mixed to obtain an added mixed signal; the added mixed signal passes through a first secondary path and is differentially mixed with the first expected signal to obtain a first residual signal; the first residual signal passes through a feedback controller to obtain a residual compensation signal, the residual compensation signal is added and mixed with the added mixed signal to form an added residual compensation signal; the added residual compensation signal is then used to pass through a second secondary path and is differentially mixed with the second expected signal to perform active noise reduction.

6. The active noise reduction method according to claim 5, characterized in that: The filter parameters corresponding to the first feedforward microphone, the second feedforward microphone and the feedback controller are optimized using the filter parameter optimization method described in claims 1-3.

7. A method for active noise reduction of headphones, characterized by: A first feedforward microphone, a second feedforward microphone and a virtual microphone are respectively set; the first feedforward microphone obtains a first signal, the second feedforward microphone obtains a second signal, and the virtual microphone obtains an expected signal; the first signal passes through a first feedforward controller and the second signal passes through a second feedforward controller and then is added and mixed to obtain an added mixed signal; the added mixed signal is differentially mixed with the expected signal through a secondary path to perform active noise reduction.

8. The active noise reduction method according to claim 7, characterized in that: The filter parameters corresponding to the first feedforward microphone and the second feedforward microphone are optimized using the filter parameter optimization method described in claim 4.

9. A filter parameter optimization system, characterized in that: It includes: signal acquisition module, signal processing module and parameter adjustment module, and the signals of each module are connected; The signal acquisition module is used to respectively set a first feedforward microphone, a second feedforward microphone, a feedback microphone and a virtual microphone; wherein the first feedforward microphone acquires a first signal, the second feedforward microphone acquires a second signal, the feedback microphone acquires a first desired signal and the virtual microphone acquires a second desired signal; The signal processing module is used to add and mix the first signal after passing through the first feedforward controller and the second signal after passing through the second feedforward controller to obtain an added mixed signal; the added mixed signal is differentially mixed with the first desired signal after passing through the first secondary path to obtain a first residual signal; the first residual signal is then passed through the feedback controller to obtain a residual compensation signal, the residual compensation signal is added and mixed with the added mixed signal to form an added residual compensation signal; the added residual compensation signal is differentially mixed with the second desired signal after passing through the second secondary path to obtain a second residual signal; The parameter adjustment module is used to adjust the filter parameters of the feedback controller according to the first residual signal; and to adjust the filter parameters of the first feedforward controller and the second feedforward controller respectively according to the second residual signal.

10. A Bluetooth headset, characterized in that Comprising: comprising: a Bluetooth headset body, wherein the Bluetooth headset body is used to implement the filter parameter optimization method described in any one of claims 1-4.

Images