Active noise reduction control system and method based on speaker

Abstract

The invention discloses an active noise reduction control system and method based on a loudspeaker, and relates to the field of loudspeaker noise reduction control. The technical scheme thereof mainly includes: using a mathematical model to construct a real-time ear canal model of a user; collecting initial noise data, calculating the amplitude value, the original phase and the quantization factor corresponding to the initial noise data at each sampling point, and obtaining compressed light noise data; designing the cutoff frequency and the weight coefficient of a hybrid noise reduction algorithm, using an improved hybrid noise reduction algorithm to process the light noise data, and obtaining an output noise-removed data set; using a sliding window to perform a sliding process of a moving average on the updated noise-removed data set, and obtaining a gain data set that is finally output; using an inverse fast Fourier transform and an accumulated differential signal to decompress and restore the gain data set, and applying the decompressed and restored signal data to a device for noise reduction, so as to realize high-efficiency and high-quality active noise reduction control of the loudspeaker.

Description

Technical Field

[0001] The present invention relates to the field of loudspeaker noise reduction control, and more particularly to a loudspeaker-based active noise reduction control system and method. Background Art

[0002] Active noise reduction (ANC) technology captures external noise through a built-in microphone and then generates opposite sound waves to cancel out the noise. Its benefits include improving sound quality, reducing ambient noise interference, and enhancing the listening experience, especially in noisy environments.

[0003] In the process of active noise reduction of the speaker, the active noise reduction system usually needs to collect environmental noise in real time and generate anti-phase sound waves to offset it. If the processing delay in the system is too large, the generated anti-phase sound waves may be out of sync with the noise, thereby reducing the noise reduction effect, and the active noise reduction algorithm usually requires higher computing resources. Especially in low-power devices (such as wireless headphones), the computing resources are limited, which may affect the noise reduction effect. In addition, as the environmental noise changes (such as fluctuations in noise frequency and intensity), the noise reduction system needs to adjust parameters quickly. If the adjustment is not timely or accurate, it may cause the noise reduction effect to be unstable, so it is necessary to design the active noise reduction control system of the speaker. Summary of the invention

[0004] In view of the problems existing in the prior art, the object of the present invention is to provide an active noise reduction control system and method based on a speaker, so as to achieve high-efficiency and high-quality noise removal in the process of active noise reduction control based on the speaker.

[0005] To achieve the above object, the present invention provides the following technical solution: an active noise reduction control system based on a speaker, the active noise reduction control system based on a speaker comprising:

[0006] Data sampling module: used to collect initial noise data and build a real-time ear canal model of the user, and perform lightweight compression on the initial noise data to obtain lightweight noise data;

[0007] Noise reduction fitting module: Use the improved hybrid noise reduction algorithm to process the light noise data, obtain the output noise-removed data set, smooth the noise-removed data set, and obtain the final output gain data set;

[0008] Output control module: used to decompress and restore the gain data set, and apply the decompressed and restored signal data to the device for noise reduction;

[0009] The modules are connected to each other via wired and / or wireless means.

[0010] Preferably, the initial noise data includes external environmental noise, ear canal residual noise and surrounding overall noise, the external environmental noise is collected by a feedforward microphone, the ear canal residual noise is collected by a feedback microphone, and the surrounding overall noise is collected by an environmental microphone.

[0011] Preferably, the method for constructing the real-time ear canal model includes: when the user uses the device for the first time, setting an initialized ear canal model, playing the o_p group of audio signals with known frequency and waveform in sequence and outputting them into the ear canal through the built-in speaker of the device, the feedback microphone in the device collects the sound wave reflection signal in the ear canal in real time, and extracts the characteristic data of the propagation and reflection of the sound wave reflection signal in the ear canal, using a mathematical model to represent the acoustic characteristics of the ear canal, using the o_p group of audio signals to continuously update the acoustic characteristics in the ear canal, and then obtaining an optimized ear canal model, which is recorded as the real-time ear canal model.

[0012] Preferably, the method for lightweight compression of the initial noise data comprises: within a period T, sampling the initial noise data P times at a fixed frequency, thereby obtaining an initial noise data set X = [x1, ..., x P ];

[0013] Calculate the difference between adjacent sampling points to get the differential signal d g =x g -x g-1 , where the initial threshold x0 is set, and all differential signals are collected to obtain the differential data set D = [d1, ..., d P ], where d h represents the hth differential signal, x e represents the initial noise data obtained by the e-th sampling, and e=1,…,g-1,g,…,P, h=1,…,g,…,P; use fast Fourier transform to process the differential data set to obtain the frequency domain signal Among them, w k represents the frequency domain signal value corresponding to the g-th differential signal, k is the index of the frequency domain, and j is the imaginary unit, j 2 = -1;

[0014] Quantize the amplitude of the frequency domain signal. Among them, |w k (g)| represents the frequency domain signal w k The quantized amplitude value, Q is the quantization factor, round() indicates rounding operation, |w k | represents the frequency domain signal w k the magnitude of

[0015] Calculate the frequency domain signal w k The original phase θ k= arg(w k ), where arg(w k ) is the frequency domain signal w k The phase angle of

[0016] The amplitude value, original phase and quantization factor corresponding to the initial noise data at each sampling point are collected to obtain compressed light noise data.

[0017] Preferably, the method for processing light noise data using an improved hybrid denoising algorithm comprises:

[0018] Standardize and normalize the light noise data, and transform the signal in the light noise data into X_f=[x_f1,…,x_f P ],x_f a represents the signal conversion representation corresponding to the a-th sample data in the light noise data, and a=1,……,P;

[0019] Applying filter H_f to the signal conversion representation X_f, the filtered signal Y_f = H_f·X_f, where the frequency response of filter H_f is designed to be f1 and f2 are the cutoff frequencies of the filter, f represents the actual frequency, and the signal Yf_1 is obtained by performing inverse Fourier transform on the filtered signal Y_f;

[0020] Perform wavelet transform on the standardized and normalized light noise data to decompose and obtain wavelet coefficients of different scales Among them, φ r,o (t) is the wavelet basis function, Xb_r,o is the wavelet coefficient at scale r and position o, and ql(t) represents the signal data at sampling point t in the light noise data;

[0021] Perform threshold processing on the wavelet coefficients to obtain the processed coefficients The wavelet coefficients after threshold processing Perform inverse wavelet transform to obtain signal Yf_2;

[0022] Combine the signal Yf_1 and the signal Yf_2 to obtain the final output signal Yf(v)=ρ1×Yf_1(v)+ρ2×Yf_2(v), where ρ1 and ρ2 are weight coefficients, Yf(v) represents the noise reduction signal output at the sampling point v, Yf_1(v) represents the signal Yf_1 at the sampling point v, and Yf_2(v) represents the signal Yf_2 at the sampling point v;

[0023] Collect all the denoised signals within the time period T to obtain the denoised data set.

[0024] Preferably, the method for obtaining the cutoff frequency of the filter includes: setting a value set, using cross-validation to obtain the best value, and applying it to the noise reduction fitting module.

[0025] Preferably, the method for obtaining the weight coefficient includes:

[0026] The particle swarm optimization algorithm is used to obtain the values ​​of weight coefficients ρ1 and ρ2, and the optimization goal is set to minimize the signal-to-noise ratio, with the constraint condition of ρ1+ρ2=1;

[0027] Treat each possible combination of ρ1 and ρ2 as a particle, randomly initialize the initial position and initial velocity of each particle, initialize the personal best position of each particle to its initial position, and initialize the global best position to the particle position with the best objective function value among all particles;

[0028] Set the particle speed update formula to:

[0029] Position update formula Among them, δ is the inertia weight, μ1 and μ2 are learning factors, τ1 and τ2 are random numbers in the range [0,1], is the position of particle i in the c+1th iteration, is the position of particle i in the cth iteration, is the velocity of particle i in the c+1th iteration, is the velocity of particle i in the cth iteration, wz_best,i is the personal best position of particle i, and qw_best is the global best position;

[0030] Calculating particle positions The corresponding target value is obtained. If the current target value is greater than the target value corresponding to the personal best position, the personal best position is updated to the current position. If the current target value is greater than the global best position target value, the global best position is updated to the current position. The termination condition is set to reach the number of iterations, and the weight coefficients ρ1 and ρ2 values ​​under the given optimization target are obtained and applied to the denoising fitting module.

[0031] Preferably, the method for smoothing the denoised data set comprises:

[0032] The first signal data and the last signal data of the denoised data set within the period time T are copied, the copied signal data are used as new critical values, and the denoised data set is updated;

[0033] Use a sliding window to perform moving average on the updated denoised data set, set the window step size to 2L+1, where L is the half step size of the window, and calculate the smoothing value for each signal data Ph(n) represents the smoothed signal data at sampling point n, Yf(n+b) represents the signal data in the sliding window of the denoised data set, and b represents a constant;

[0034] Collect the smoothed signal data at all sampling points to obtain the final output gain data set.

[0035] Preferably, the method for decompressing and restoring the gain data set comprises:

[0036] The inverse fast Fourier transform is used to decode each signal data in the gain data set, and then the signal data is restored by accumulating the differential signal to obtain the decompressed and restored signal data within the cycle time T.

[0037] A speaker-based active noise reduction control method, comprising:

[0038] Step S1: constructing a real-time ear canal model of the user using a mathematical model;

[0039] Step S2: Collect initial noise data, calculate the amplitude value, original phase and quantization factor corresponding to the initial noise data at each sampling point, and obtain compressed light noise data;

[0040] Step S3: designing the cutoff frequency and weight coefficient of the hybrid denoising algorithm, using the improved hybrid denoising algorithm to process the light noise data, and obtaining the output denoised data set;

[0041] Step S4: Use a sliding window to perform a sliding process of moving average on the updated noise-removed data set to obtain a gain data set as a final output;

[0042] Step S5: decompress and restore the gain data set using inverse fast Fourier transform and accumulated differential signals, and apply the decompressed and restored signal data to the device for noise reduction.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] In the present invention, by constructing a real-time ear canal model of the user, the system can make personalized adjustments to the ear canal characteristics of each user, thereby significantly improving the noise reduction effect. Differential pulse code modulation (DPCM) and fast Fourier transform (FFT) are combined to compress noise data, effectively reducing the amount of data while retaining key signal information, reducing the burden of calculation and storage. The improved hybrid noise reduction algorithm is adopted, combined with a variety of noise reduction methods, which can better handle different noise characteristics, improve noise reduction performance, and maintain signal quality. The weight coefficient of the noise reduction algorithm is adjusted using a particle swarm optimization algorithm, which can globally optimize the noise reduction performance and adapt to different noise environments and user needs. The noise-removed data set is smoothed by the moving average of the sliding window, which reduces the random fluctuation of the noise and improves the stability and comfort of the final noise reduction signal. While reducing noise, the system tries to maintain the original characteristics of the signal, reduces distortion in the noise reduction process, makes the audio content clearer and more realistic, ensures that the system runs at low power consumption, and improves the overall efficiency and battery life of the system. The adaptive control and optimization mechanism continuously improves the overall efficiency of intelligent analysis, updates the ear canal model and optimizes the noise reduction algorithm in real time to adapt to changes in different users and environments, thereby providing a continuous and stable noise reduction effect. Multiple modules work together to achieve comprehensive data collection, processing and analysis, provide real-time and accurate analysis of detection data, and realize optimal planning of transmission nodes in the process of active noise reduction control based on speakers, thereby improving the efficiency of data transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The present invention proposes a structural schematic diagram of an active noise reduction control system based on a speaker;

[0046] Figure 2 Schematic diagram of the method steps applied to the speaker-based active noise reduction control system in the present invention. DETAILED DESCRIPTION

[0047] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0048] Embodiment 1

[0049] Reference Figure 1 and Figure 2 Embodiment 1 further illustrates an active noise reduction control system based on a speaker proposed by the present invention.

[0050] An architecture design is performed for an active noise reduction control system based on a speaker, wherein the hardware architecture includes a speaker, a multi-microphone array, a digital signal processor (DSP), a controller, and a power management module. The speaker is used to play noise reduction signals and audio content. The multi-microphone array includes a feedforward microphone, a feedback microphone, and an ambient microphone, which are respectively used to collect ambient noise, speaker output sound waves, and noise in the user's ear canal. A dedicated DSP chip of a digital signal processor (DSP) is used for real-time audio signal processing, including the execution of a noise reduction algorithm. A microcontroller or an embedded processor is used for overall control of the system, including parameter adjustment, data communication, etc. The power management module is responsible for the power supply and management of the system to ensure low power consumption operation.

[0051] Data sampling module: used to collect initial noise data and build a real-time ear canal model of the user, and perform lightweight compression on the initial noise data to obtain lightweight noise data;

[0052] The initial noise data includes external environmental noise, residual ear canal noise and overall surrounding noise. The external environmental noise is collected by a feedforward microphone, the residual ear canal noise is collected by a feedback microphone, and the overall surrounding noise (the overall noise situation of the surrounding environment) is collected by an environmental microphone.

[0053] The method for constructing a real-time ear canal model includes: when a user uses the device for the first time, setting an initialization ear canal model, and playing the audio signals of the o_p group with known frequency and waveform in sequence into the ear canal through the built-in speaker of the device (usually a sweep signal or a pulse signal). The feedback microphone in the device collects the sound wave reflection signals in the ear canal in real time. These signals include the influence of the shape, size, curvature and other characteristics of the ear canal on the sound waves, and extracts the characteristic data of the propagation and reflection of the sound wave reflection signals in the ear canal. A mathematical model is used to represent the acoustic characteristics of the ear canal. Commonly used models include FIR (finite impulse response) filters, IIR (infinite impulse response) filters or more complex acoustic transmission line models. The o_p group of audio signals are used to continuously update the acoustic characteristics in the ear canal (such as the position of the earphone, slight changes in the shape of the ear canal), and then the optimized ear canal model is obtained, which is recorded as the real-time ear canal model.

[0054] By building a user's ear canal model, the noise reduction effect is further optimized. Since the ear canal shape of each user is different, a personalized ear canal model can improve the noise reduction effect while avoiding unnecessary waste of computing resources, thereby achieving better power consumption control.

[0055] Methods for lightweight compression of initial noise data include:

[0056] In the period T, the initial noise data is sampled P times at a fixed frequency, and then the initial noise data set X = [x1, ..., x P ];

[0057] Calculate the difference between adjacent sampling points to get the differential signal d g =x g -x g-1 , where the initial threshold x0 is set, and all differential signals are collected to obtain the differential data set D = [d1, ..., d P ], where d h represents the hth differential signal, x e represents the initial noise data obtained by the e-th sampling, and e=1,…,g-1,g,…,P, h=1,…,g,…,P; use fast Fourier transform to process the differential data set to obtain the frequency domain signal Among them, w k represents the frequency domain signal value corresponding to the g-th differential signal, k is the index of the frequency domain, and j is the imaginary unit, j 2 = -1;

[0058] Quantize the amplitude of the frequency domain signal to reduce the amount of data. Among them, |w k (g)| represents the frequency domain signal w k The quantized amplitude value, Q is the quantization factor, which determines the quantization accuracy, round() indicates rounding operation, |w k | represents the frequency domain signal w k the magnitude of

[0059] Calculate the frequency domain signal w k The original phase θ k = arg(w k ), where arg(w k ) is the frequency domain signal w k The phase angle of

[0060] The amplitude value, original phase and quantization factor corresponding to the initial noise data at each sampling point are collected to obtain compressed light noise data.

[0061] Using differential pulse code modulation to calculate the difference between adjacent sampling points, DPCM reduces the amount of data by reducing the dynamic range of the signal and provides a more compact signal expression for subsequent processing; using fast Fourier transform (FFT) for quantization, FFT converts the time domain signal into the frequency domain signal, retaining only the important frequency components, and further reduces the amount of data through quantization. The two are used in combination to further improve data compression efficiency and provide greater flexibility in noise reduction and signal processing. This combined method can apply compression technology in the time domain and frequency domain respectively, thereby maximizing the effect of data compression at different signal processing stages while maintaining signal quality as much as possible.

[0062] Noise reduction fitting module: Use the improved hybrid noise reduction algorithm to process the light noise data, obtain the output noise-removed data set, smooth the noise-removed data set, and obtain the final output gain data set;

[0063] The method of using the improved hybrid denoising algorithm to process light noise data includes:

[0064] Standardize and normalize the light noise data, and transform the signal in the light noise data into X_f=[x_f1,…,x_f P ],x_f a represents the signal conversion representation corresponding to the a-th sample data in the light noise data, and a=1,……,P;

[0065] Applying filter H_f to the signal conversion representation X_f, the filtered signal Y_f = H_f·X_f, where the frequency response of filter H_f is designed to be f1 and f2 are the cutoff frequencies of the filter, and the filtered signal Y_f is subjected to inverse Fourier transform to obtain the signal Yf_1;

[0066] Perform wavelet transform on the standardized and normalized light noise data to decompose and obtain wavelet coefficients of different scales Among them, φ r,o (t) is the wavelet basis function, Xb_r,o is the wavelet coefficient at scale r and position o, and ql(t) represents the signal data at sampling point t in the light noise data;

[0067] Perform threshold processing on the wavelet coefficients to obtain the processed coefficients The wavelet coefficients after threshold processing Perform inverse wavelet transform to obtain signal Yf_2;

[0068] Combine the signal Yf_1 and the signal Yf_2 to obtain the final output signal Yf(v)=ρ1×Yf_1(v)+ρ2×Yf_2(v), where ρ1 and ρ2 are weight coefficients, Yf(v) represents the noise reduction signal output at the sampling point v, Yf_1(v) represents the signal Yf_1 at the sampling point v, and Yf_2(v) represents the signal Yf_2 at the sampling point v;

[0069] Collect all the denoised signals within the time period T to obtain the denoised data set.

[0070] The hybrid denoising algorithm combines multiple denoising methods to achieve better results when processing signals with different noise characteristics. By combining multiple methods, the original characteristics of the signal can be better preserved while denoising, signal distortion can be reduced, and data latency can be reduced. In addition, by adjusting the weights or parameters of each method, it can flexibly adapt to different application scenarios and needs.

[0071] The method for obtaining the cutoff frequencies f1 and f2 of the filter includes: setting a value set, using cross-validation to obtain the best value, and applying it to the noise reduction fitting module.

[0072] The method for obtaining the weight coefficients ρ1 and ρ2 includes:

[0073] The particle swarm optimization algorithm is used to obtain the values ​​of weight coefficients ρ1 and ρ2, and the optimization goal is set to minimize the signal-to-noise ratio, with the constraint condition of ρ1+ρ2=1;

[0074] Treat each possible combination of ρ1 and ρ2 as a particle, randomly initialize the initial position and initial velocity of each particle, initialize the personal best position of each particle to its initial position, and initialize the global best position to the particle position with the best objective function value among all particles;

[0075] Set the particle speed update formula to:

[0076]

[0077] Position update formula Among them, δ is the inertia weight, which controls the influence of the particle's historical velocity on the current velocity. μ1 and μ2 are learning factors, which control the degree to which the particle learns its own optimal position and the global optimal position, respectively. τ1 and τ2 are random numbers in the range of [0,1], which are used to introduce randomness. is the position of particle i in the c+1th iteration, is the position of particle i in the cth iteration, is the velocity of particle i in the c+1th iteration, is the velocity of particle i in the cth iteration, wz_best,i is the personal best position of particle i, and qw_best is the global best position;

[0078] Calculating particle positions The corresponding target value is obtained. If the current target value is greater than the target value corresponding to the personal best position, the personal best position is updated to the current position. If the current target value is greater than the global best position target value, the global best position is updated to the current position. The termination condition is set to reach the number of iterations, and the weight coefficients ρ1 and ρ2 values ​​under the given optimization target are obtained and applied to the denoising fitting module.

[0079] The weight coefficients of the hybrid denoising algorithm using particle swarm optimization have the advantages of global optimization capability, ease of use, strong adaptability, efficient computing performance and wide applicability. This makes the particle swarm optimization algorithm an effective tool for optimizing the weight coefficients in the denoising algorithm, and can provide high-quality optimization results when processing complex signals.

[0080] Methods for smoothing a denoised dataset include:

[0081] The first signal data and the last signal data of the denoised data set within the period time T are copied, the copied signal data are used as new critical values, and the denoised data set is updated;

[0082] Use a sliding window to perform moving average on the updated denoised data set, and set the step size of the window to 2L+1, where L is the half step size of the window. The step size of the window is designed based on experimental data analysis or experience. For each signal data, calculate the smoothing value Ph(n) represents the smoothed signal data at sampling point n, Yf(n+b) represents the signal data in the sliding window of the denoised data set, and b represents a constant;

[0083] Collect the smoothed signal data at all sampling points to obtain the final output gain data set.

[0084] The moving average is used to locally average each signal to reduce the impact of random noise and thus smooth the signal.

[0085] Output control module: used to decompress and restore the gain data set, and apply the decompressed and restored signal data to the device for noise reduction;

[0086] Methods for decompressing and restoring the gain data set include:

[0087] The inverse fast Fourier transform is used to decode each signal data in the gain data set, and then the signal data is restored by accumulating the differential signal to obtain the decompressed and restored signal data within the cycle time T.

[0088] Through the above steps, the final restored signal data is very close to the original signal, especially when the noise reduction effect is significant, it is the same as the original signal.

[0089] Embodiment 2

[0090] Reference Figure 1 and Figure 2 , Embodiment 2 further illustrates an active noise reduction control system based on a speaker proposed by the present invention.

[0091] A method for active noise reduction control based on a speaker, comprising the following steps:

[0092] Step S1: constructing a real-time ear canal model of the user using a mathematical model;

[0093] Step S2: Collect initial noise data, calculate the amplitude value, original phase and quantization factor corresponding to the initial noise data at each sampling point, and obtain compressed light noise data;

[0094] Step S3: designing the cutoff frequency and weight coefficient of the hybrid denoising algorithm, using the improved hybrid denoising algorithm to process the light noise data, and obtaining the output denoised data set;

[0095] Step S4: Use a sliding window to perform a sliding process of moving average on the updated noise-removed data set to obtain a gain data set as a final output;

[0096] Step S5: decompress and restore the gain data set using inverse fast Fourier transform and accumulated differential signals, and apply the decompressed and restored signal data to the device for noise reduction.

[0097] An active noise reduction control system based on a loudspeaker, applied to an active noise reduction control method based on a loudspeaker, includes:

[0098] Data sampling module: used to collect initial noise data and build a real-time ear canal model of the user, and perform lightweight compression on the initial noise data to obtain lightweight noise data;

[0099] Noise reduction fitting module: Use the improved hybrid noise reduction algorithm to process the light noise data, obtain the output noise-removed data set, smooth the noise-removed data set, and obtain the final output gain data set;

[0100] Output control module: used to decompress and restore the gain data set, and apply the decompressed and restored signal data to the device for noise reduction;

[0101] The modules are connected to each other via wired and / or wireless means.

[0102] In addition, according to the implementation of the present application, the process described in the figure of an active noise reduction control system based on a speaker can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application. Of course, the architecture shown in the figure of an active noise reduction control system based on a speaker is only exemplary. When implementing different devices, adaptive selection or adjustment can be performed according to actual needs.

[0103] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0104] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. An active noise reduction control system based on a speaker, characterized in that: The speaker-based active noise reduction control system includes: Data sampling module: used to collect initial noise data and build a real-time ear canal model of the user, and perform lightweight compression on the initial noise data to obtain lightweight noise data; The method for lightweight compression of initial noise data comprises: In the period T, the initial noise data is sampled P times at a fixed frequency, and then the initial noise data set X = [x1, ..., x P ]; Calculate the difference between adjacent sampling points to get the differential signal d g =x g -x g-1 , where the initial threshold x0 is set, and all differential signals are collected to obtain the differential data set D = [d1, ..., d P ], where d h represents the hth differential signal, x e represents the initial noise data obtained by the e-th sampling, and e=1,…,g-1,g,…,P, h=1,…,g,…,P; use fast Fourier transform to process the differential data set to obtain the frequency domain signal Among them, w k represents the frequency domain signal value corresponding to the g-th differential signal, k is the index of the frequency domain, and j is the imaginary unit, j 2 = -1; Quantize the amplitude of the frequency domain signal. Among them, |w k (g)| represents the frequency domain signal w k The quantized amplitude value, Q is the quantization factor, round() indicates rounding operation, |w k | represents the frequency domain signal w k the magnitude of Calculate the frequency domain signal w k The original phase θ k = arg(w k ), where arg(w k ) is the frequency domain signal w k The phase angle of Collect the amplitude value, original phase and quantization factor corresponding to the initial noise data at each sampling point to obtain compressed light noise data; Noise reduction fitting module: Use the improved hybrid noise reduction algorithm to process the light noise data, obtain the output noise-removed data set, smooth the noise-removed data set, and obtain the final output gain data set; The method for processing light noise data using an improved hybrid denoising algorithm comprises: Standardize and normalize the light noise data, and transform the signal in the light noise data into X_f=[x_f1,…,x_f P ],x_f a represents the signal conversion representation corresponding to the a-th sample data in the light noise data, and a=1,……,P; Applying filter H_f to the signal conversion representation X_f, the filtered signal Y_f = H_f·X_f, where the frequency response of filter H_f is designed to be f1 and f2 are the cutoff frequencies of the filter, f represents the actual frequency, and the signal Yf_1 is obtained by performing inverse Fourier transform on the filtered signal Y_f; Perform wavelet transform on the standardized and normalized light noise data to decompose and obtain wavelet coefficients of different scales Among them, φ r,o (t) is the wavelet basis function, Xb_r,o is the wavelet coefficient at scale r and position o, and ql(t) represents the signal data at sampling point t in the light noise data; Perform threshold processing on the wavelet coefficients to obtain the processed coefficients The wavelet coefficients after threshold processing Perform inverse wavelet transform to obtain signal Yf_2; Combine the signal Yf_1 and the signal Yf_2 to obtain the final output signal Yf(v)=ρ1×Yf_1(v)+ρ2×Yf_2(v), where ρ1 and ρ2 are weight coefficients, Yf(v) represents the noise reduction signal output at the sampling point v, Yf_1(v) represents the signal Yf_1 at the sampling point v, and Yf_2(v) represents the signal Yf_2 at the sampling point v; Collect all denoised signals within a time period T to obtain a denoised data set; The method for smoothing the denoised data set comprises: The first signal data and the last signal data of the denoised data set within the period time T are copied, the copied signal data are used as new critical values, and the denoised data set is updated; Use a sliding window to perform moving average on the updated denoised data set, set the window step size to 2L+1, where L is the half step size of the window, and calculate the smoothing value for each signal data Ph(n) represents the smoothed signal data at sampling point n, Yf(n+b) represents the signal data in the sliding window of the denoised data set, and b represents a constant; Collect the smoothed signal data at all sampling points to obtain the final output gain data set; Output control module: used to decompress and restore the gain data set, and apply the decompressed and restored signal data to the device for noise reduction; The modules are connected to each other via wired and / or wireless means.

2. The active noise reduction control system based on a speaker according to claim 1, characterized in that: The initial noise data includes external environmental noise, ear canal residual noise and surrounding overall noise. The external environmental noise is collected by a feedforward microphone, the ear canal residual noise is collected by a feedback microphone, and the surrounding overall noise is collected by an environmental microphone.

3. The active noise reduction control system based on a speaker according to claim 2, characterized in that: The method for constructing the real-time ear canal model includes: when a user uses the device for the first time, setting an initialization ear canal model, playing the o_p group of audio signals with known frequencies and waveforms in sequence and outputting them into the ear canal through the built-in speaker of the device, the feedback microphone in the device collects the sound wave reflection signal in the ear canal in real time, and extracts the characteristic data of the propagation and reflection of the sound wave reflection signal in the ear canal, using a mathematical model to represent the acoustic characteristics of the ear canal, using the o_p group of audio signals to continuously update the acoustic characteristics in the ear canal, and then obtaining an optimized ear canal model, which is recorded as the real-time ear canal model.

4. The active noise reduction control system based on a speaker according to claim 3, characterized in that: The method for obtaining the cutoff frequency of the filter includes: setting a value set, using cross-validation to obtain the best value, and applying it to a noise reduction fitting module.

5. The active noise reduction control system based on a speaker according to claim 4, characterized in that: The method for obtaining the weight coefficient includes: The particle swarm optimization algorithm is used to obtain the values ​​of weight coefficients ρ1 and ρ2, and the optimization goal is set to minimize the signal-to-noise ratio, with the constraint condition of ρ1+ρ2=1; Treat each possible combination of ρ1 and ρ2 as a particle, randomly initialize the initial position and initial velocity of each particle, initialize the personal best position of each particle to its initial position, and initialize the global best position to the particle position with the best objective function value among all particles; Set the particle speed update formula to: Position update formula Among them, δ is the inertia weight, μ1 and μ2 are learning factors, τ1 and τ2 are random numbers in the range [0,1], is the position of particle i in the c+1th iteration, is the position of particle i in the cth iteration, is the velocity of particle i in the c+1th iteration, is the velocity of particle i in the cth iteration, wz_best,i is the personal best position of particle i, and qw_best is the global best position; Calculate the particle position wz_ i c+1 The corresponding target value is obtained. If the current target value is greater than the target value corresponding to the personal best position, the personal best position is updated to the current position. If the current target value is greater than the global best position target value, the global best position is updated to the current position. The termination condition is set to reach the number of iterations, and the weight coefficients ρ1 and ρ2 values ​​under the given optimization target are obtained and applied to the denoising fitting module.

6. The active noise reduction control system based on a speaker according to claim 5, characterized in that: The method for decompressing and restoring the gain data set comprises: The inverse fast Fourier transform is used to decode each signal data in the gain data set, and then the signal data is restored by accumulating the differential signal to obtain the decompressed and restored signal data within the cycle time T.

7. A speaker-based active noise reduction control method, applied to a speaker-based active noise reduction control system according to any one of claims 1 to 6, characterized in that: include: Step S1: constructing a real-time ear canal model of the user using a mathematical model; Step S2: Collect initial noise data, calculate the amplitude value, original phase and quantization factor corresponding to the initial noise data at each sampling point, and obtain compressed light noise data; Step S3: designing the cutoff frequency and weight coefficient of the hybrid denoising algorithm, using the improved hybrid denoising algorithm to process the light noise data, and obtaining the output denoised data set; Step S4: Use a sliding window to perform a sliding process of moving average on the updated noise-removed data set to obtain a gain data set as a final output; Step S5: decompress and restore the gain data set using inverse fast Fourier transform and accumulated differential signals, and apply the decompressed and restored signal data to the device for noise reduction.

Images