Active noise reduction method and device of head-mounted wireless earphone and head-mounted wireless earphone

Abstract

The invention relates to the technical field of earphone noise reduction, and discloses an active noise reduction method and device for a head-mounted wireless earphone and the head-mounted wireless earphone. The method comprises the following steps of: performing time division multiplexing sampling through double microphone arrays of left and right acoustic cavities of the earphone to obtain four paths of microphone array sampling data and performing correlation function beam forming calculation to generate sound source direction distribution data; calculating a geometric acoustic path difference of the earphone, and establishing a direction delay parameter table corresponding to each effective sound source direction; driving a multi-channel adaptive filter bank according to the direction delay parameter table, performing directional filtering, and generating a directional inversion signal group; and taking the intensity information in the sound source direction distribution data as a weighting coefficient to perform weighted synthesis on the directional inversion signal group, forming a comprehensive inversion signal, applying a feedforward delay correction value, and outputting a self-adaptive noise reduction signal. According to the method, continuous and stable noise reduction performance can be kept in a complex dynamic multi-sound-source environment, and the auditory experience of a user is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of headphone noise reduction, and in particular to an active noise reduction method and device for a wireless headphone, and the wireless headphone. Background Art

[0002] Traditional active noise cancellation systems primarily utilize feedforward or feedback noise cancellation architectures, uniformly processing ambient noise through a single, global delay compensation algorithm. However, in real-world headphone wear environments, users often encounter complex noise sources from multiple directions, including traffic noise in front, conversations to the sides, and mechanical equipment noise behind. These sound sources from different directions have different spectral characteristics and propagation paths.

[0003] The main problem with existing technologies is that they cannot effectively address the need for noise reduction in environments with multi-directional sound sources. Traditional single delay compensation algorithms use fixed delay parameters to uniformly process noise from all directions, ignoring the differences in geometric paths and phase characteristics of sound sources from different directions reaching the ears. Especially in the headphone wearing environment, due to the blocking effect of the head, sound sources from the side and rear will produce significant phase delay changes during the propagation process. Traditional algorithms are unable to accurately compensate for this direction-related phase characteristic, resulting in the noise source from a specific direction not only being unable to be effectively suppressed, but may even produce an enhancement effect due to delay matching errors. Summary of the Invention

[0004] The main purpose of the present invention is to provide an active noise reduction method, device and head-mounted wireless headphones. The present invention can maintain continuous and stable noise reduction performance in a complex dynamic multi-sound source environment, significantly improving the user's auditory experience.

[0005] To achieve the above object, the present invention provides an active noise reduction method for a wireless headset, comprising the following steps: Time-division multiplexing sampling is performed through the dual microphone arrays in the left and right acoustic cavities of the headset to obtain four-channel microphone array sampling data and perform correlation function beamforming calculations to generate sound source direction distribution data; Calculating the headphone geometric acoustic path difference using the direction information in the sound source direction distribution data, and establishing a directional delay parameter table corresponding to each effective sound source direction; Drive a multi-channel adaptive filter group according to the directional delay parameter table to perform directional filtering on the four-channel microphone array sampling data to generate a directional inverted signal group; The intensity information in the sound source direction distribution data is used as a weighting coefficient to perform weighted synthesis on the directional inverted signal group to form a comprehensive inverted signal and apply a feedforward delay correction amount to output an adaptive noise reduction signal.

[0006] Optionally, in a first implementation of the first aspect of the present invention, the dual-microphone arrays of the left and right acoustic cavities of the earphones perform time-division multiplexing sampling to obtain four-channel microphone array sampling data and perform correlation function beamforming calculations to generate sound source direction distribution data, including: A first front-facing microphone and a first rear-facing microphone are provided in the left acoustic cavity of the earphone, and a second front-facing microphone and a second rear-facing microphone are provided in the right acoustic cavity of the earphone, thereby obtaining a dual-microphone array; Establishing a time division multiplexing control sequence with the first front microphone, the second front microphone, the first rear microphone, and the second rear microphone in the dual-microphone array as a cyclic order; Driving the dual-microphone array in sequence according to the time division multiplexing control sequence to perform acoustic signal acquisition and analog-to-digital conversion to obtain digital sampling signals, and storing the digital sampling signals in a circular buffer according to microphone position identifiers to organize and form four-channel microphone array sampling data; A correlation function beamforming calculation is performed on the four-channel microphone array sampling data to generate sound source direction distribution data.

[0007] Optionally, in a second implementation of the first aspect of the present invention, performing correlation function beamforming calculation on the four-channel microphone array sampling data to generate sound source direction distribution data includes: Performing fast Fourier transform on the four-channel microphone array sampling data to obtain a left ear front microphone frequency domain signal, a right ear front microphone frequency domain signal, a left ear rear microphone frequency domain signal, and a right ear rear microphone frequency domain signal; Constructing a forward cross-correlation function using the left ear forward microphone frequency domain signal and the right ear forward microphone frequency domain signal, and constructing a backward cross-correlation function using the left ear backward microphone frequency domain signal and the right ear backward microphone frequency domain signal; Driving the forward cross-correlation function and the backward cross-correlation function to perform a scanning beamforming operation according to a preset angular interval to obtain a power spectral density value; The power spectrum peak detection algorithm is used to identify the angles of each effective sound source that exceeds the noise floor threshold, and the sound source direction distribution data containing direction information and intensity information is constructed based on the corresponding power spectrum density values.

[0008] Optionally, in a third implementation of the first aspect of the present invention, calculating the headphone geometric acoustic path difference using the directional information in the sound source direction distribution data and establishing a directional delay parameter table corresponding to each effective sound source direction includes: A spherical coordinate system is established with the center of the head as the origin, and the center points of the headset's microphone array are set as the negative and positive horizontal axis coordinate positions, respectively, to form the headset's geometric acoustic coordinate reference; Extracting the direction information of each effective sound source from the sound source direction distribution data, and calculating the difference in direct acoustic path length between the left and right ears corresponding to each sound source angle according to the headphone geometric acoustic coordinate reference; Calculating a head shadow effect correction coefficient based on a cosine function value of an angle of each effective sound source, and performing path difference compensation correction for lateral and rearward sound sources on the direct acoustic path length difference value to obtain a compensated and corrected path length difference value; The compensated path length difference is divided by the sound velocity constant to convert it into a propagation time delay difference, and the directional delay parameter table is constructed in combination with the frequency segment parameters.

[0009] Optionally, in a fourth implementation of the first aspect of the present invention, driving a multi-channel adaptive filter group according to the directional delay parameter table to perform directional filtering on the four-channel microphone array sampling data to generate a directional inverted signal group includes: Constructing a corresponding adaptive finite impulse response filter for each sound source direction based on the delay time parameters of each effective sound source direction in the direction delay parameter table; Inputting the four-channel microphone array sampling data into an adaptive finite impulse response filter to perform minimum mean square error calculation to obtain an error signal, and updating the coefficient parameters of each adaptive finite impulse response filter in real time according to the error signal and the input signal after delay compensation to obtain a multi-channel adaptive filter bank; Using the phase delay information corresponding to each effective sound source direction in the directional delay parameter table to drive a multi-channel adaptive filter group, perform cascade phase compensation, and obtain a phase-corrected filter group output signal; The output signal of the filter group after the phase correction is subjected to inverse phase transformation and automatic gain control adjustment to generate a direction-dependent inverse phase signal group.

[0010] Optionally, in a fifth implementation of the first aspect of the present invention, the step of using the phase delay information corresponding to each effective sound source direction in the directional delay parameter table to drive a multi-channel adaptive filter group, performing cascade phase compensation, and obtaining a phase-corrected filter group output signal includes: Extract the phase delay value corresponding to each effective sound source direction from the direction delay parameter table, and calculate the phase delay distribution of each effective sound source direction at different frequency points in combination with the frequency segmentation parameter; generating all-pass filter coefficients corresponding to each effective sound source direction based on the phase delay distribution, and constructing an all-pass filter bank matching the number of sound source directions; cascade-connecting the all-pass filter bank and the corresponding adaptive finite impulse response filters in the multi-channel adaptive filter bank according to a direction matching principle to form a cascade filter bank architecture; The cascaded filter bank architecture is driven to perform a joint process of amplitude filtering and phase correction on an input signal, and output a filter bank output signal after phase correction.

[0011] Optionally, in a sixth implementation of the first aspect of the present invention, using the intensity information in the sound source direction distribution data as a weighting coefficient to perform weighted synthesis on the directional inverted signal group to form a comprehensive inverted signal and applying a feedforward delay correction amount to output an adaptive noise reduction signal includes: Extracting intensity information corresponding to each effective sound source direction from the sound source direction distribution data, and calculating a weighting coefficient corresponding to each effective sound source direction through a normalization operation; Based on the angle difference between the directions of each effective sound source, the cosine function is used to calculate and construct the inter-directional interference suppression matrix. When the angle between adjacent directions is less than the preset threshold, the corresponding interference suppression factor is activated. Applying the weighting coefficient and the interference suppression factor to the direction-specific inverted signal group simultaneously to perform a weighted linear combination operation to generate an initial inverted signal; Applying a time domain window function smoothing process to the initial inverted signal and establishing a dynamic power monitoring mechanism to perform adaptive gain adjustment, and outputting a comprehensive inverted signal; The sound source direction change information in the continuous time window is monitored and a feedforward delay correction amount is applied to the integrated inverted phase signal to output an adaptive noise reduction signal.

[0012] Optionally, in a seventh implementation of the first aspect of the present invention, monitoring the sound source direction change information within the continuous time window and applying a feedforward delay correction amount to the integrated inverted signal to output an adaptive noise reduction signal includes: Calculating the angular velocity value of each effective sound source based on the sound source direction change information of the sound source direction distribution data within the continuous time window, and generating the sound source position prediction data at the next moment through a first-order prediction filter operation; Inputting the sound source position prediction data into a Kalman filter to optimize the prediction accuracy to obtain a predicted angle change, and calculating a feedforward delay correction amount based on the predicted angle change and a partial derivative of the delay with respect to the angle; The direction change amplitude of the dominant sound source within adjacent sampling periods is monitored and a direction switching detection mechanism is established. When the direction change exceeds a preset angle threshold and the duration meets the switching condition, an exponential decay function is used to perform interpolation transition on the filter coefficients to obtain the interpolation transition processing result; Dynamically update the delay parameters of the integrated inverted signal according to the feedforward delay correction amount and the interpolation transition processing result, and output an adaptive noise reduction signal.

[0013] The present invention also provides an active noise reduction device for a wireless headset, comprising: A calculation module is used to perform time-division multiplexing sampling through the dual microphone arrays in the left and right acoustic cavities of the headset, obtain sampling data from the four-channel microphone array, and perform correlation function beamforming calculations to generate sound source direction distribution data; An establishment module is used to calculate the headphone geometric acoustic path difference using the direction information in the sound source direction distribution data, and establish a directional delay parameter table corresponding to each effective sound source direction; A directional filtering module is used to drive a multi-channel adaptive filter group according to the directional delay parameter table, perform directional filtering on the four-channel microphone array sampling data, and generate a directional anti-phase signal group; The output module is used to perform weighted synthesis on the directional inverted signal group using the intensity information in the sound source direction distribution data as a weighting coefficient to form a comprehensive inverted signal and apply a feedforward delay correction amount to output an adaptive noise reduction signal.

[0014] The present invention also provides a wireless headset for implementing the steps of any of the above methods.

[0015] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of any of the above methods are implemented.

[0016] In summary, the technical solution provided by the present invention utilizes a time-division multiplexing sampling mechanism to avoid electromagnetic interference and signal crosstalk between the four microphones in traditional synchronous sampling methods, significantly reducing crosstalk between adjacent directions. An improved correlation function beamforming algorithm, combined with four-microphone cross-correlation calculation and angle-dependent phase compensation, effectively eliminates the negative impact of head occlusion on lateral and rearward sound source detection accuracy, achieving accurate sound source localization from 0 to 360 degrees. A directional delay calculation model based on the geometric acoustic characteristics of the earphones is established to provide personalized delay compensation for the specific propagation path and phase characteristics of each sound source direction, fundamentally overcoming the technical deficiency of traditional technologies in which a single delay parameter cannot adapt to the differences in multi-directional sound sources. A parallel processing architecture of a multi-channel adaptive filter bank is used to independently generate an anti-phase signal with precise phase compensation for each detected sound source direction, resolving the technical problem of traditional single filters being unable to simultaneously process sound sources from multiple directions. A sound source intensity-weighted synthesis algorithm and an inter-directional interference suppression mechanism are established, and cross-directional suppression is achieved through an interference suppression matrix, effectively avoiding phase interference and signal conflict between multiple anti-phase signals and ensuring coordinated operation of the anti-phase signals from all directions. By using a feedforward delay correction and smooth transition algorithm, a dynamic tracking mechanism based on sound source motion prediction is established. This mechanism proactively adjusts delay compensation parameters and employs exponential decay transition processing, effectively resolving the issue of interrupted noise reduction effects in traditional passive tracking methods when the sound source position changes. This invention maintains continuous and stable noise reduction performance in complex, dynamic, multi-source environments, significantly improving the user's auditory experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 1 is a schematic diagram of the steps of an active noise reduction method for a wireless headphone according to an embodiment of the present invention; Figure 2 FIG. 4 is a structural block diagram of an active noise reduction device for a wireless headphone according to an embodiment of the present invention.

[0018] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0020] Reference Figure 1 This embodiment provides an active noise reduction method for a wireless headset, comprising the following steps: S1, through the dual microphone array of the left and right acoustic cavities of the headset, time-division multiplexing sampling is performed to obtain four-channel microphone array sampling data and perform correlation function beamforming calculation to generate sound source direction distribution data; The physical structure of the wireless headset incorporates microphones in the left and right acoustic cavities. The left acoustic cavity houses a first front-facing microphone and a first rear-facing microphone, while the right acoustic cavity houses a second front-facing microphone and a second rear-facing microphone. This creates a dual-microphone array system capable of spatial directional distribution perception. In the hardware driver logic design, a time-division multiplexing control sequence is established, with the first front-facing microphone, the second front-facing microphone, the first rear-facing microphone, and the second rear-facing microphone as the sampling order. This sequence activates the four microphones for sampling at predetermined intervals. This control sequence allows for independent acquisition of four channels of audio signals within each complete sampling cycle. The analog acoustic signals are converted to digital signals using a 16-bit analog-to-digital converter, generating digital sample data. All acquired digital sample signals are then temporally sorted by microphone location and stored in corresponding circular buffers in a predefined order. This results in a data stream with four channels: ML1, MR1, ML2, and MR2. Each sampling channel maintains independent and traceable temporal consistency, forming a four-channel microphone array digital sample data set. Based on this dataset, correlation function beamforming is performed. Fast Fourier transforms are introduced to convert the time-domain sampled data into a frequency-domain representation. A cross-correlation enhancement function with directional scanning capabilities is constructed. Complex multiplication is used to superimpose the frequency components of each channel, and phase compensation is introduced to improve the ability to suppress non-ideal effects such as head occlusion. During the directional scanning process, possible sound source directions are traversed within an angular range of 0 to 360 degrees in steps of 5 degrees. For each direction, the cross-correlation energy value is calculated. After the scan is complete, the power spectral density results for all directions are normalized and peak extracted, forming a sound source direction distribution dataset containing sound source angle and corresponding intensity information.

[0021] S2, calculating the headphone geometric acoustic path difference using the direction information in the sound source direction distribution data, and establishing a directional delay parameter table corresponding to each effective sound source direction; Specifically, the headset and head structure are abstracted at the spatial modeling level. A spherical coordinate system is established with the center of the user's head as the origin. The centers of the two microphone arrays of the headset are mapped to two spatial points symmetrically distributed along the horizontal axis. The center of the left ear microphone array is set to the negative horizontal axis position coordinate (-0.085 meters, 0, 0), and the center of the right ear microphone array is set to the positive horizontal axis position coordinate (0.085 meters, 0, 0). This forms a headset acoustic coordinate reference system with clear geometric boundaries and directional determination criteria. The direction information of all valid sound sources identified in the sound source direction distribution data is extracted one by one and used as the independent variable. The straight-line distance difference between each sound source direction angle and the center points of the left and right microphone arrays is analyzed in the spherical coordinate system. The specific path difference value is obtained by applying trigonometric functions to the geometric projection relationship between the left-right distance between the headset and the sound source incident angle. Its basic form is to multiply the distance between the two microphone centers by the sine of the corresponding sound source angle to obtain the path length difference in the uncompensated state. Because the human head is not a perfect sphere and has a certain degree of occlusion and reflection on sound propagation, a head shadow correction factor related to the directional angle is introduced into the path difference. This factor is calculated based on the cosine function of the sound source angle and varies nonlinearly with the deviation angle of the sound source from the front. It is multiplied by the path difference to compensate for the propagation path of sound sources located to the side or rear, effectively compensating for path difference deviations caused by factors such as ear obstruction, head occlusion, and diffraction. The compensated path length difference is divided by the sound velocity constant to obtain the propagation time delay difference in the corresponding direction—the time difference between the sound source signal reaching one microphone and the other. Based on this, the delay parameters are segmented and managed in conjunction with the frequency response characteristics. A frequency segmented parameter model is established at 1 / 3 octave intervals across the entire effective noise reduction band from 200Hz to 8kHz. Each frequency segment is assigned a specific delay compensation value, thus refining the directional delay difference into a multi-band, multi-angle composite representation. Each of the above-mentioned sound source direction angles, the corresponding propagation delay time difference, and the phase offset of each frequency band are organized together into a direction delay parameter table.

[0022] S3, driving a multi-channel adaptive filter group according to the directional delay parameter table, performing directional filtering on the four-channel microphone array sampling data, and generating a directional anti-phase signal group; It should be noted that, based on the propagation time delay values ​​recorded for each valid sound source direction in the directional delay parameter table, a set of finite impulse response filters is configured for that direction, serving as the core computational unit for targeted noise reduction. These filters are instantiated for each sound source direction, forming a parallel processing network with directionality independence and structural consistency. Each adaptive finite impulse response filter uses the delay time parameters provided in the directional delay table as its primary input to control the filter's delay compensation structure and initialize the filter's tap length and sampling window size, enabling precise phase matching and amplitude cancellation of the target direction's sound source signal. After the filter is activated, the four-channel microphone array sample data is synchronously input into the adaptive finite impulse response filter. Based on the directional mapping results, each filter selects the channel combination most relevant to its target direction as input. Based on this input, the filter weights are adaptively updated in real time using a minimum mean square error (MMSE) algorithm. The error signal is generated by differentially calculating the current filter output with a preset reference signal. This error is then multiplied by the delay-compensated input signal, thereby continuously optimizing the filter coefficient vector. This adaptive mechanism ensures that the filter corresponding to each direction can dynamically adapt to changes in the input sound source, achieving continuous noise suppression in the selected direction. Considering that sound source propagation is accompanied not only by time delay but also by frequency-dependent phase shift, the amplitude response of each filter is updated, incorporating the phase delay information specified in the directional delay parameter table to further perform cascade phase compensation on the filter output. This compensation process is achieved by introducing a complex all-pass filter structure or a delay rotation vector to ensure that the signals from each direction are phase-aligned before superposition or synthesis. After processing, the phase-corrected filter bank output signal is obtained. The phase-corrected filter bank output signal is then inverted, rotating the signal's phase by 180 degrees, thereby achieving noise interference reduction when superimposed on the original input channel. Furthermore, considering that the energy level of the original noise signal fluctuates over time and direction, an automatic gain control module is introduced to maintain the equality and synchronization of the inverted signals in the amplitude dimension, providing closed-loop feedback adjustment of the output amplitude of each signal. This adjustment process monitors the power ratio between the inverted signal and the target noise in real time and adjusts the output amplitude within a preset gain range, ultimately generating a set of inverted signals for each direction.

[0023] S4, performing weighted synthesis on the directional inverted signal group using the intensity information in the sound source direction distribution data as a weighting coefficient to form a comprehensive inverted signal and applying a feedforward delay correction amount to output an adaptive noise reduction signal.

[0024] Specifically, the intensity information corresponding to each valid sound source direction is extracted from the direction-intensity mapping matrix. This intensity information, expressed as a power spectral density, represents the energy significance of each sound source in the sound field. To ensure that the contributions of multiple sound sources are accurately and proportionally added during the synthesis process, these intensity values ​​are normalized. The intensity of each direction is divided by the sum of the intensities of all valid sound sources. This generates a set of normalized weighting coefficients, which sum to 1 and control energy conservation and directional distribution. The relative spatial relationship between multiple sound sources is analyzed, focusing on the angular difference between adjacent directions. Based on this, an interference suppression mechanism is introduced to prevent phase interference when signals from these directions are superimposed. A cosine function is used to calculate the angle between each pair of valid directions, and an inter-directional interference suppression matrix is ​​constructed. The matrix elements are defined as functions that depend on the angular difference. When the angle between any two directions is less than a set threshold (e.g., 30 degrees), the corresponding matrix elements are assigned an attenuation weight, which is used to introduce a directional isolation factor into the signal synthesis process. This factor mechanism effectively suppresses nonlinear cross-interference between multiple sound source signals by dynamically adjusting the synthesis weights between adjacent directions. The normalized weighting coefficient and the inter-directional interference suppression factor are simultaneously applied to the generated directional anti-phase signal group, and all directional signals are fused through weighted linear combination calculation to generate a preliminary integrated anti-phase signal. In this process, the energy dominance of the dominant sound source direction is maintained, and the noise reduction efficiency loss caused by the adjacent interference direction is effectively reduced, so that the synthetic signal has higher spatial specificity and phase consistency. The initial anti-phase signal is smoothed by applying a time domain window function, and the time domain window function is introduced to smooth the transition of the signal boundary to prevent the energy mutation caused by frame sampling from affecting the subjective auditory quality; and a dynamic power monitoring mechanism is established to monitor the power ratio between the current anti-phase signal and the original input signal in real time. When the energy difference between the two exceeds the threshold, the gain adjustment unit is automatically triggered to adjust the overall gain level so that the amplitude of the anti-phase signal is always in an energy matching state with the target noise. In order to improve the adaptive ability of the system in dynamic environments, a feedforward delay correction mechanism is introduced. That is, after monitoring the change of the sound source direction within the continuous time window, the future position of the sound source is predicted according to the angular velocity estimation model, and the corresponding delay correction amount is applied to the current integrated inverted signal in combination with the path delay change trend. In this way, the dynamic switching of the delay compensation parameters is completed in advance when the sound source is displaced, avoiding the interruption or failure of noise reduction caused by sudden changes in direction, and outputting an adaptive noise reduction signal.

[0025] In one example, the dual microphone arrays of the left and right acoustic cavities of the earphones perform time division multiplexing sampling to obtain four-channel microphone array sampling data and perform correlation function beamforming calculation to generate sound source direction distribution data, including: A first front-facing microphone and a first rear-facing microphone are provided in the left acoustic cavity of the earphone, and a second front-facing microphone and a second rear-facing microphone are provided in the right acoustic cavity of the earphone, thereby obtaining a dual-microphone array; Establishing a time division multiplexing control sequence with the first front microphone, the second front microphone, the first rear microphone, and the second rear microphone in the dual-microphone array as a cyclic order; Driving the dual-microphone array in sequence according to the time division multiplexing control sequence to perform acoustic signal acquisition and analog-to-digital conversion to obtain digital sampling signals, and storing the digital sampling signals in a circular buffer according to microphone position identifiers to organize and form four-channel microphone array sampling data; A correlation function beamforming calculation is performed on the four-channel microphone array sampling data to generate sound source direction distribution data.

[0026] In this example, a specific microphone distribution strategy is designed based on the headphone's structural characteristics and ear-worn configuration. Two microphones with fixed physical positions and complementary directivities are placed in the left acoustic cavity of the headphone. One microphone is located in the area facing the user's front, acting as the first front-facing microphone, and the other is located in the area facing the user's back, acting as the first rear-facing microphone. Similarly, two microphones are placed in the right acoustic cavity at corresponding positions: a second front-facing microphone and a second rear-facing microphone. This mirror-symmetric structure forms a dual-microphone array with spatial coverage. To ensure multi-channel audio data capture without introducing electromagnetic crosstalk and power supply jitter issues caused by simultaneous multi-channel sampling, the system uses time-division multiplexing to uniformly schedule the four microphone channels, constructing a sampling control sequence. This control sequence is progressively cycled in the order of "first front-facing microphone → second front-facing microphone → first rear-facing microphone → second rear-facing microphone." Only one microphone is activated during each sampling cycle, with a fixed sampling duration of 31.25 microseconds. The sampling cycle for all four channels is controlled within 500 microseconds, ensuring the system's overall cyclic stability and temporal resolution. Simultaneously, the sampling controller synchronizes the drive signal with the analog-to-digital conversion circuit, ensuring that each activated microphone immediately acquires analog sound signals within the sampling window. These signals are then converted to digital signals using a high-precision 16-bit ADC (analog-to-digital converter). The converted digital samples are labeled according to the physical location of the microphones: ML1 (front left), MR1 (front right), ML2 (rear left), and MR2 (rear right). These signals are then sequentially written into a circular buffer with a capacity of 2048 sampling points, generating structured digital sample data for the four-channel microphone array. Correlation function beamforming is then applied to the four-channel microphone array sample data, enabling multi-directional sound source detection and distribution analysis based on this technology. A fast Fourier transform (FFT) is then performed on each channel's sample data, with a transform length of 512 points, to extract frequency domain information and enhance analysis of complex sound sources. This conversion results in a frequency-domain complex signal. The system constructs an improved cross-correlation function to detect sound source responses in different directions in space. This function contains conventional inter-channel product terms and introduces a phase compensation factor corresponding to the theoretical delay to ensure that rearward sound sources can still be accurately detected under the premise of head occlusion effect. The construction of the cross-correlation function is based on the following strategy: the complex conjugate product of the left front and right front channels is used as the forward cross-correlation term, and the complex conjugate product of the left rear and right rear channels is used as the backward cross-correlation term, and each is multiplied by a phase correction factor derived from the angle θ to form a composite angular response expression. The direction is traversed in steps of 5 degrees within the range of 0 to 360 degrees. Each direction corresponds to a theoretical delay difference, and a set of directional angle-cross-correlation response function values ​​is constructed from this traversal.In each direction, the power spectrum density of the cross-correlation function moduli at different frequency points is superimposed to form the directional response intensity. A power spectrum peak detection algorithm is then used to identify all locations where the signal intensity is significantly higher than the background noise threshold (set to 10dB higher). These directional angles and corresponding power intensities are then saved as part of the direction-intensity mapping data. The directions of all identified valid sound sources and their corresponding power information are combined to form a directional distribution data structure. This data structure is based on 72 directional distribution grid points, with one grid every 5 degrees, completely covering the 360-degree sound source scanning area in the entire space. The data is updated 20 times per second, enabling real-time capture and modeling of the dynamic characteristics of sound sources in multi-directional noise fields.

[0027] In one example, performing correlation function beamforming calculation on the four-channel microphone array sampling data to generate sound source direction distribution data includes: Performing fast Fourier transform on the four-channel microphone array sampling data to obtain a left ear front microphone frequency domain signal, a right ear front microphone frequency domain signal, a left ear rear microphone frequency domain signal, and a right ear rear microphone frequency domain signal; Constructing a forward cross-correlation function using the left ear forward microphone frequency domain signal and the right ear forward microphone frequency domain signal, and constructing a backward cross-correlation function using the left ear backward microphone frequency domain signal and the right ear backward microphone frequency domain signal; Driving the forward cross-correlation function and the backward cross-correlation function to perform a scanning beamforming operation according to a preset angular interval to obtain a power spectral density value; The power spectrum peak detection algorithm is used to identify the angles of each effective sound source that exceeds the noise floor threshold, and the sound source direction distribution data containing direction information and intensity information is constructed based on the corresponding power spectrum density values.

[0028] In this example, frequency domain conversion is performed on data from a four-channel microphone array. A 512-point data window is extracted from each microphone's sampling buffer. After windowing, it is input into the Fast Fourier Transform (FFT) module to obtain frequency-domain complex signals for the left ear forward, right ear forward, left ear backward, and right ear backward channels. Spatial cross-correlation modeling is performed on these four frequency-domain signals. The left ear forward signal is combined with the right ear forward signal to construct a forward cross-correlation function, while the left ear backward signal is combined with the right ear backward signal to construct a backward cross-correlation function. These cross-correlation functions are considered directional response functions that are spatially sensitive to specific angles. They are integrated based on the directional angle and frequency characteristics, and a phase compensation factor is added to eliminate the phase offset caused by the varying propagation path lengths of the sound source, thereby enhancing the response to sound sources at different angles. In the directional space dimension, the system sets a scanning step size of 5 degrees, traversing a total of 72 angular directions from 0 to 360 degrees. In each direction, the forward and backward cross-correlation functions are configured as the theoretical propagation conditions for that direction. The complex outputs within the frequency domain are amplitude-integrated, with the integration band set between 200 Hz and 8 kHz to cover the frequency bandwidth most sensitive to human hearing. The integration result for each direction represents the power spectral density value at that angle, reflecting the energy of the sound source in that direction. After completing the full-angle scan, a 72-dimensional power spectral density distribution sequence is generated. A peak detection algorithm is used to traverse this distribution sequence and analyze all directions that are significantly above the background noise threshold. The threshold is set at 10 dB above the noise floor to ensure that only valid sound sources with significant energy advantage are identified. The detection process combines directional continuity with power variation trends to determine whether abnormal peaks have real physical significance, filtering out isolated sudden changes or background jitter interference. For each identified valid sound source direction, its angle information and corresponding power spectral density value are recorded, which together form the direction-intensity mapping data. This mapping data is organized into a directional distribution matrix, where each row represents a sound source direction, including its spatial angle and corresponding sound intensity. This matrix, a 72×1-dimensional matrix, represents the omnidirectional detection results at every 5-degree interval. This directional distribution matrix is ​​updated in real time at a rate of 20 times per second to ensure the system's high time-domain responsiveness, ultimately generating sound source directional distribution data.

[0029] In one example, calculating the headphone geometric acoustic path difference using the direction information in the sound source direction distribution data and establishing a direction delay parameter table corresponding to each effective sound source direction includes: A spherical coordinate system is established with the center of the head as the origin, and the center points of the headset's microphone array are set as the negative and positive horizontal axis coordinate positions, respectively, to form the headset's geometric acoustic coordinate reference; Extracting the direction information of each effective sound source from the sound source direction distribution data, and calculating the difference in direct acoustic path length between the left and right ears corresponding to each sound source angle according to the headphone geometric acoustic coordinate reference; Calculating a head shadow effect correction coefficient based on a cosine function value of an angle of each effective sound source, and performing path difference compensation correction for lateral and rearward sound sources on the direct acoustic path length difference value to obtain a compensated and corrected path length difference value; The compensated path length difference is divided by the sound velocity constant to convert it into a propagation time delay difference, and the directional delay parameter table is constructed in combination with the frequency segment parameters.

[0030] In this example, a mathematical expression is established for the geometric relationship between the headset and the head at the spatial modeling level. The center point of the human head wearing the headset is used as the origin of the spherical coordinate system to construct a three-dimensional spatial coordinate reference system. The horizontal left and right axes are used as the reference axes of the headset structure. The geometric center point of the left ear microphone array is set at the negative position of the horizontal axis with a coordinate of -0.085 meters, and the geometric center point of the right ear microphone array is set at the positive position of the horizontal axis with a coordinate of +0.085 meters, forming an acoustic reference reference with the head as the center and symmetrical distribution. All valid sound source direction angle information within the current time window is extracted from the sound source direction distribution data. Each angle represents the azimuth offset of the sound source relative to the front of the user's head, in degrees, ranging from 0 to 360 degrees. Based on the geometric relationship between the sound source angle and the lateral reference, the system calculates the difference in path lengths between the left and right microphones for that sound source in a free field. This difference, under ideal free-field conditions, is calculated by multiplying the distance between the earphone centers by the sine function of the sound source's direction angle. This distance is set to 0.17 meters, the symmetrical distance between the left and right microphone arrays. This method yields the theoretical difference in path lengths between the left and right ears for each sound source, representing the difference in the direct sound wave propagation path. This difference directly corresponds to the time phase difference received between the ears. Because the human head is not a perfectly point-symmetrical object, its geometric contours exhibit complex physical properties such as protrusions, obstructions, and reflections. Especially for lateral and rearward sound sources, the shadowing effects of the pinna, skull, and neck on the sound waves can significantly alter the original path propagation structure. Therefore, the system incorporates a head shadowing correction mechanism into path length difference modeling. This mechanism constructs a correction coefficient model using the cosine value of the sound source's directional angle as a variable. This correction coefficient approaches 1 when directly in front of the sound source and significantly increases when 90 degrees to the side or 180 degrees to the rear. The system defaults to a cosine function of the angle, calculated to account for symmetry and reflect the physical laws of sound field diffraction attenuation. The introduction of the correction coefficient effectively compensates for the original path length difference in areas affected by the head structure, making the final model more realistic for acoustic propagation. Applying this correction coefficient to weightedly correct the path length difference yields a set of effective path difference values ​​reflecting actual propagation conditions. This path difference exists in positive and negative directions, representing the actual physical quantity of sound arriving at one earphone microphone earlier or later, respectively. Based on the compensated path length difference, the path length difference is converted into a propagation time difference to construct the time parameters used for filter phase control and delay correction. This conversion process uses a constant sound velocity as a scaling factor, set at 343 meters per second. This ratio conversion converts the length into a time delay, measured in microseconds. This delay describes the time inconsistency caused by angular differences between the sound waves from one microphone to another. To ensure frequency-domain adaptability of the delay compensation parameters, the system incorporates a frequency segmentation modeling strategy.Within the 200Hz to 8kHz frequency range, the system constructs a segmented structure in 1 / 3 octave units. For each frequency point within this segment, an independent delay correction value is established and mapped to the directional angle, thus forming a two-dimensional delay parameter table covering all directional angles and all frequency ranges. This delay parameter table includes directional angles, path difference correction values, propagation delay values, and phase response data for different frequency bands. The entire table covers a total of 72 directions, with complete frequency band parameters for each direction.

[0031] In one example, the driving of the multi-channel adaptive filter group according to the directional delay parameter table to perform directional filtering on the four-channel microphone array sampling data to generate a directional inverted signal group includes: Constructing a corresponding adaptive finite impulse response filter for each sound source direction based on the delay time parameters of each effective sound source direction in the direction delay parameter table; Inputting the four-channel microphone array sampling data into an adaptive finite impulse response filter to perform minimum mean square error calculation to obtain an error signal, and updating the coefficient parameters of each adaptive finite impulse response filter in real time according to the error signal and the input signal after delay compensation to obtain a multi-channel adaptive filter bank; Using the phase delay information corresponding to each effective sound source direction in the directional delay parameter table to drive a multi-channel adaptive filter group, perform cascade phase compensation, and obtain a phase-corrected filter group output signal; The output signal of the filter group after the phase correction is subjected to inverse phase transformation and automatic gain control adjustment to generate a direction-dependent inverse phase signal group.

[0032] In this example, the delay parameters for each valid sound source direction in the directional delay parameter table are fully annotated, along with the directional angle, frequency band, and path compensation value. Based on these delay parameters, an adaptive finite impulse response filter with a fixed structure is instantiated for each direction. This filter, based on an FIR structure, has a fixed order (e.g., 128) and a coefficient matrix that can be adjusted in real time. Each filter unit is responsible for constructing the inverted signal from that direction and mapping it to the delay time in the parameter table to ensure that each directional signal is processed independently. The system extracts raw audio signal segments from the four-channel microphone array sampling data and inputs them into the FIR filters for each direction in a time-aligned manner. During this process, the system applies a compensatory delay to the input signal based on the propagation delay parameters for the corresponding direction. This delay is performed before the input filter to align the signal with the target sound source on the time axis. During processing, the filter continuously performs a minimum mean square error (MMSE) calculation, extracting the difference between the current filter output and the target reverse interference reference signal as an error signal. The filter coefficients are then adjusted in real time based on the product of this error signal and the input signal. This update process occurs continuously, reassessing the residual magnitude between the current input's time-domain characteristics and the target reverse response within each sampling period, thereby adaptively approximating the coefficients towards the optimal direction. Because each filter channel operates independently based on a different direction, this mechanism is executed in parallel, ultimately forming a multi-channel adaptive filter bank, where each channel is independently modeled and adjusted for a specific sound source direction and maintained synchronously within the system. To ensure that the filter bank's output signal maintains full phase alignment with the actual noise source in the frequency domain, the system applies additional phase compensation to each filter output, incorporating phase delay information contained in the directional delay parameter table. This process is achieved by introducing a phase adjustment structure after each filter channel. This structure can be viewed as a cascaded phase compensator, whose task is to precisely rotate the output signal while maintaining the amplitude characteristics, achieving a perfect phase reversal with the target noise signal at the corresponding frequency point. This type of phase compensation not only accounts for the phase shift caused by the ideal propagation path, but also incorporates the frequency response shifts caused by head shadowing and actual geometric asymmetries. This ensures that the system maintains stable phase accuracy even in highly complex and dynamically changing sound field environments. The phase-corrected filter bank output signal undergoes an inverse phase transformation and automatic gain control adjustment.The polarity of the current output signal is reversed, allowing it to effectively cancel out interference with the original noise signal during the physical superposition process. Automatic gain control is then implemented. This strategy dynamically adjusts the output gain of the inverted signal within each time period by detecting the energy ratio between the inverted signal and the input sound source signal. The gain range is set between -20dB and +10dB, ensuring that the reverse path signal strength remains within the matching range as the ambient noise energy increases or decreases. Through these steps, a set of directional inverted signals is ultimately generated.

[0033] In one example, the step of using the phase delay information corresponding to each effective sound source direction in the directional delay parameter table to drive a multi-channel adaptive filter group, performing cascade phase compensation, and obtaining a phase-corrected filter group output signal includes: Extract the phase delay value corresponding to each effective sound source direction from the direction delay parameter table, and calculate the phase delay distribution of each effective sound source direction at different frequency points in combination with the frequency segmentation parameter; generating all-pass filter coefficients corresponding to each effective sound source direction based on the phase delay distribution, and constructing an all-pass filter bank matching the number of sound source directions; cascade-connecting the all-pass filter bank and the corresponding adaptive finite impulse response filters in the multi-channel adaptive filter bank according to a direction matching principle to form a cascade filter bank architecture; The cascaded filter bank architecture is driven to perform a joint process of amplitude filtering and phase correction on an input signal, and output a filter bank output signal after phase correction.

[0034] In this example, the phase delay values ​​corresponding to each valid direction are extracted from the directional delay parameter table one by one, indexed by the sound source angle. Combined with a frequency segmentation strategy, a segmented interpolation calculation is performed on the representative frequencies within each frequency band within the effective human hearing range of 200Hz to 8kHz, at intervals of 1 / 3 octave, to obtain the phase delay distribution data for the sound source direction at all frequency points. This step expands the original single phase delay into a multi-band phase response vector, ensuring that the filter structure can accurately simulate the phase characteristics caused by actual sound source propagation at different frequency points and addressing the bandwidth limitations of single-frequency compensation. Based on the above phase delay distribution results, an all-pass filter with a phase delay characteristic matching the corresponding direction is generated for each valid sound source direction. The design principle of the all-pass filter is to maintain the amplitude characteristics of the input signal unchanged and only control its phase distribution so that the phase response of the output signal at each frequency point can fully correspond to the target delay. To achieve this goal, the system calculates a set of all-pass filter parameters for each direction based on the frequency-domain variation of phase delay. These parameters form the core control factors of the filter, driving a multi-order recursive structure within the implementation architecture to achieve precise phase rotation. Because the all-pass filters themselves do not affect signal energy, they are used in tandem with the main channel adaptive filters without introducing new gain imbalances or nonlinear distortion. The system constructs an all-pass filter bank that exactly matches the number of sound source directions using a one-to-one mapping relationship. Each filter path is configured with a separate all-pass filter instance, which is cascaded with the corresponding directional channel in the multi-channel adaptive finite impulse response filter to form a direction-specific cascaded filter subsystem. This structure, known as a cascaded filter bank architecture, combines FIR filters with adjustable amplitude response and all-pass filters with adjustable phase response, achieving joint optimization of both amplitude adjustment and phase correction. This cascaded filter bank architecture processes the four-microphone array sampled signal on a channel-by-channel basis for each frame. During the processing, the FIR component first performs amplitude filtering and dynamic coefficient updates based on the aforementioned minimum mean square error criterion. The all-pass component then performs phase correction on the same signal under frequency band distribution, ensuring that each output has reverse characteristics in the waveform, directional phase synchronization in the frequency domain, and propagation delay compensation capabilities in the time domain. The system uses a control structure module to uniformly manage the parameter update process, input-output synchronization mechanism, and directional channel switching strategy in the cascade structure, ensuring that each channel maintains data integrity and parameter continuity even when the sound source position changes and the number of directions increases or decreases dynamically. The output signal processed by this architecture is the output of the filter bank after phase correction.

[0035] In one example, the step of using the intensity information in the sound source direction distribution data as a weighting coefficient to perform weighted synthesis on the directional inverted signal group to form a comprehensive inverted signal and applying a feedforward delay correction amount to output an adaptive noise reduction signal includes: Extracting intensity information corresponding to each effective sound source direction from the sound source direction distribution data, and calculating a weighting coefficient corresponding to each effective sound source direction through a normalization operation; Based on the angle difference between the directions of each effective sound source, the cosine function is used to calculate and construct the inter-directional interference suppression matrix. When the angle between adjacent directions is less than the preset threshold, the corresponding interference suppression factor is activated. Applying the weighting coefficient and the interference suppression factor to the direction-specific inverted signal group simultaneously to perform a weighted linear combination operation to generate an initial inverted signal; Applying a time domain window function smoothing process to the initial inverted signal and establishing a dynamic power monitoring mechanism to perform adaptive gain adjustment, and outputting a comprehensive inverted signal; The sound source direction change information in the continuous time window is monitored and a feedforward delay correction amount is applied to the integrated inverted phase signal to output an adaptive noise reduction signal.

[0036] In this example, the directional angle and corresponding intensity value of each effective sound source are extracted from the front-end sound source direction distribution data. The intensity information is expressed in the form of power spectral density, covering a directional range of 0 to 360 degrees, and a sound field mapping set of 72 directional units is constructed at every 5-degree interval. The system normalizes the sum of the intensities of all effective sound sources using the power values ​​marked in the data structure, and divides the intensity of each sound source by the total intensity value to obtain a normalized weighting coefficient. While calculating the weighting coefficient, an inter-directional interference suppression mechanism is introduced to solve the phase overlap and mis-cancellation problems that occur when multiple adjacent directional sound sources are superimposed on opposite-phase signals. This mechanism uses the angle difference as an input variable, calculates the angle between any two effective directions, and introduces a cosine function to express the suppression weight. When the angle is less than a preset threshold (such as 30 degrees), the corresponding suppression factor is activated to form a symmetrical inter-directional interference suppression matrix. Each element in the matrix corresponds to the suppression strength between two directions. When the angle is close to or approaches zero, the suppression strength tends to be maximum. Therefore, the system automatically reduces the mutual influence of the anti-phase signals in adjacent directions during directional fusion, thereby realizing a control strategy that enhances the multi-source spatial filtering capability and minimizes interference. The weighting coefficient and interference suppression factor are applied to the anti-phase signal group in each direction at the same time, and all directional channel signals are proportionally fused through a weighted linear combination method. On each signal channel, the corresponding weighting coefficient is multiplied by the directional signal strength and the interaction factor in the directional interference suppression matrix, thereby realizing an anti-phase signal synthesis strategy centered on the dominant sound source and measuring weight attenuation according to spatial divergence. The result of this operation is the initial anti-phase signal, which reflects the contribution relationship of each direction in the amplitude dimension, retains the anti-phase characteristics in the phase dimension, and completes cross-interference suppression in the spatial dimension. To prevent abrupt changes in the temporal continuity of the synthesized antiphase signal, amplitude jumps, or residual multi-directional phase interference, a time-domain window function is applied to the signal after synthesis. A 64-point Hamming window or cosine window is used for convolution and smoothing, resulting in a gradual transition at the waveform boundaries in the time domain. This ensures a smooth and continuous phase curve for the antiphase signal, thus preventing unstable reflections in the sound field. The system also incorporates a dynamic power monitoring mechanism during this stage. This mechanism calculates the ratio of the total power of the initial antiphase signal to the noise power of the original microphone array input signal in real time, and sets an upper threshold for the power ratio (e.g., 0.95) as a control boundary. If the antiphase signal energy is detected to be excessive, approaching or exceeding the energy of the original noise signal, the system automatically activates a gain compression algorithm to reduce the overall gain of the antiphase signal to ensure it remains within a controllable amplitude range. If the antiphase signal energy is low, the system activates a gain boost mechanism to replenish the energy without distortion, thereby maintaining optimal interference cancellation. This process results in a synthesized antiphase signal.The system continuously monitors the directional distribution of sound sources within multiple continuous time windows and extracts the rate of angular change of each direction across different time windows. By calculating the angular velocity, it derives the directional change trend, which is then used as input for feedforward prediction and correction of the incoming filter system delay parameters. Based on the predicted value, the system pre-calculates the propagation delay correction corresponding to the directional angle at the next moment and adds this correction to the delay path of the integrated inverted signal, achieving parameter-level forward scheduling and outputting an adaptive noise reduction signal.

[0037] In one example, monitoring the sound source direction change information within the continuous time window and applying a feedforward delay correction amount to the integrated inverted signal to output an adaptive noise reduction signal includes: Calculating the angular velocity value of each effective sound source based on the sound source direction change information of the sound source direction distribution data within the continuous time window, and generating the sound source position prediction data at the next moment through a first-order prediction filter operation; Inputting the sound source position prediction data into a Kalman filter to optimize the prediction accuracy to obtain a predicted angle change, and calculating a feedforward delay correction amount based on the predicted angle change and a partial derivative of the delay with respect to the angle; The direction change amplitude of the dominant sound source within adjacent sampling periods is monitored and a direction switching detection mechanism is established. When the direction change exceeds a preset angle threshold and the duration meets the switching condition, an exponential decay function is used to perform interpolation transition on the filter coefficients to obtain the interpolation transition processing result; Dynamically update the delay parameters of the integrated inverted signal according to the feedforward delay correction amount and the interpolation transition processing result, and output an adaptive noise reduction signal.

[0038] In this example, the system's sound source direction distribution data, which is updated 20 times per second, is slidingly integrated to extract valid sound source direction records from the three most recent time segments at any given moment. The angular information for each sound source direction at different time points forms its angular sequence on the time axis. By dividing the angular difference between the current and previous time points by the sampling interval (set to 50 milliseconds), the angular velocity of each sound source direction—the rate of change of the angular direction per unit time—is calculated. This angular velocity information is input into a first-order predictive filter to estimate the future position. The prediction is based on the current sound source angle plus the product of the angular velocity and the time interval, forming the new angle of the sound source at the next moment. This first-order filter structure is highly responsive to uniform changes in speed over short periods of time and is suitable for dynamic sound field scenarios, such as slow panning of sounds and turning heads to speak. Given that sound source changes in real environments are not always smooth and linear and are subject to uncertainties such as noise interference, directional misjudgment, and background disturbances, to improve the accuracy and robustness of the predicted values, after completing the initial prediction, the predicted angles are input into a Kalman filter for multivariable state optimization. In this scenario, the Kalman filter uses the angle as the state variable, the current observation angle of the system as the observation value, and sets the process noise variance to 0.1 degrees. 2 , the observation noise variance is 0.5 degrees 2 The system recursively integrates the prior prediction with the current observation to output the predicted angle change. The predicted angle change is multiplied by the angle partial derivative with respect to the preset delay to obtain the expected impact of the current directional angle change on the propagation delay, i.e., the feedforward delay correction. Furthermore, to prevent frequent directional jitter or background interference from misleading judgment, the system implements a direction switch detection mechanism. This mechanism continuously monitors the angle changes of the dominant sound source between adjacent sampling periods. It sets a sudden change threshold (e.g., 15 degrees) as the change limit and a switch duration threshold (e.g., meeting the sudden change condition for three consecutive periods) as the judgment basis. When a direction change exceeds the threshold and the duration meets the condition, the system determines that the dominant sound source has experienced a true direction switch. After a direction switch is determined, the system initiates a filter interpolation transition mechanism to prevent output signal jumps caused by sudden changes in filter parameters. This transition mechanism uses an exponential decay function to construct an interpolation curve, with the current filter coefficient and the target filter coefficient to be updated as the boundary. The function is defined as controlling the mixing ratio of the new and old coefficients at each transition sampling point by an exponential factor. The system sets a smoothing time constant of 10 milliseconds to control the transition speed and stability. The final interpolation result is used as the update coefficient for the filter in the current cycle, achieving a smooth transition from the old filter state to the new filter state. The delay parameters of the integrated inverted signal are dynamically updated based on the feedforward delay correction and the interpolation transition results. The current operating parameter set of the filter is reset to ensure that the inverted signal output is always in the optimal delay alignment state, ultimately outputting an adaptive noise reduction signal.

[0039] Reference Figure 2 This embodiment provides an active noise reduction device for a wireless headset, comprising: Computing module 1 is used to perform time-division multiplexing sampling through the dual microphone arrays in the left and right acoustic cavities of the headset, obtain four-channel microphone array sampling data, and perform correlation function beamforming calculations to generate sound source direction distribution data; Establishing module 2, for calculating the headphone geometric acoustic path difference using the direction information in the sound source direction distribution data, and establishing a directional delay parameter table corresponding to each effective sound source direction; A directional filtering module 3 is used to drive a multi-channel adaptive filter group according to the directional delay parameter table, perform directional filtering on the four-channel microphone array sampling data, and generate a directional anti-phase signal group; The output module 4 is used to perform weighted synthesis on the directional inverted signal group using the intensity information in the sound source direction distribution data as a weighting coefficient to form a comprehensive inverted signal and apply a feedforward delay correction amount to output an adaptive noise reduction signal.

[0040] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0041] An embodiment of the present invention further provides a wireless headset, which is used to implement the above method.

[0042] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, apparatus, article, or method comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, apparatus, article, or method. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, apparatus, article, or method comprising the element.

[0043] The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An active noise reduction method for a wireless headset, characterized in that: include: Time-division multiplexing sampling is performed through the dual microphone arrays in the left and right acoustic cavities of the headset to obtain four-channel microphone array sampling data and perform correlation function beamforming calculations to generate sound source direction distribution data; Calculating the headphone geometric acoustic path difference using the direction information in the sound source direction distribution data, and establishing a directional delay parameter table corresponding to each effective sound source direction; Drive a multi-channel adaptive filter group according to the directional delay parameter table to perform directional filtering on the four-channel microphone array sampling data to generate a directional inverted signal group; The intensity information in the sound source direction distribution data is used as a weighting coefficient to perform weighted synthesis on the directional inverted signal group to form a comprehensive inverted signal and apply a feedforward delay correction amount to output an adaptive noise reduction signal.

2. The active noise reduction method for a wireless headset according to claim 1, wherein: The method performs time-division multiplexing sampling through the dual microphone arrays of the left and right acoustic cavities of the earphone, obtains four-channel microphone array sampling data, and performs correlation function beamforming calculation to generate sound source direction distribution data, including: A first front-facing microphone and a first rear-facing microphone are provided in the left acoustic cavity of the earphone, and a second front-facing microphone and a second rear-facing microphone are provided in the right acoustic cavity of the earphone, thereby obtaining a dual-microphone array; Establishing a time division multiplexing control sequence with the first front microphone, the second front microphone, the first rear microphone, and the second rear microphone in the dual-microphone array as a cyclic order; driving the dual-microphone array in sequence according to the time-division multiplexing control sequence to perform acoustic signal acquisition and analog-to-digital conversion to obtain digital sampling signals, and storing the digital sampling signals in a circular buffer according to microphone position identifiers to form four-channel microphone array sampling data; A correlation function beamforming calculation is performed on the four-channel microphone array sampling data to generate sound source direction distribution data.

3. The active noise reduction method for wireless headphones according to claim 1, wherein: The performing correlation function beamforming calculation on the four-channel microphone array sampling data to generate sound source direction distribution data includes: Performing fast Fourier transform on the four-channel microphone array sampling data to obtain a left ear front microphone frequency domain signal, a right ear front microphone frequency domain signal, a left ear rear microphone frequency domain signal, and a right ear rear microphone frequency domain signal; Constructing a forward cross-correlation function using the left ear forward microphone frequency domain signal and the right ear forward microphone frequency domain signal, and constructing a backward cross-correlation function using the left ear backward microphone frequency domain signal and the right ear backward microphone frequency domain signal; Driving the forward cross-correlation function and the backward cross-correlation function to perform a scanning beamforming operation according to a preset angular interval to obtain a power spectral density value; The power spectrum peak detection algorithm is used to identify the angles of each effective sound source that exceeds the noise floor threshold, and the sound source direction distribution data containing direction information and intensity information is constructed based on the corresponding power spectrum density values.

4. The active noise reduction method for wireless headphones according to claim 1, wherein: The calculating the headphone geometric acoustic path difference by using the direction information in the sound source direction distribution data and establishing a direction delay parameter table corresponding to each effective sound source direction includes: A spherical coordinate system is established with the center of the head as the origin, and the center points of the headset's microphone array are set as the negative and positive horizontal axis coordinate positions, respectively, to form the headset's geometric acoustic coordinate reference; Extracting the direction information of each effective sound source from the sound source direction distribution data, and calculating the difference in direct acoustic path length between the left and right ears corresponding to each sound source angle according to the headphone geometric acoustic coordinate reference; Calculating a head shadow effect correction coefficient based on a cosine function value of an angle of each effective sound source, and performing path difference compensation correction for lateral and rearward sound sources on the direct acoustic path length difference value to obtain a compensated and corrected path length difference value; The compensated path length difference is divided by the sound velocity constant to convert it into a propagation time delay difference, and the directional delay parameter table is constructed in combination with the frequency segment parameters.

5. The active noise reduction method for wireless headphones according to claim 1, wherein: The method of driving a multi-channel adaptive filter group according to the directional delay parameter table, performing directional filtering on the four-channel microphone array sampling data, and generating a directional inverted signal group includes: Constructing a corresponding adaptive finite impulse response filter for each sound source direction based on the delay time parameters of each effective sound source direction in the direction delay parameter table; Inputting the four-channel microphone array sampling data into an adaptive finite impulse response filter to perform minimum mean square error calculation to obtain an error signal, and updating the coefficient parameters of each adaptive finite impulse response filter in real time according to the error signal and the input signal after delay compensation to obtain a multi-channel adaptive filter bank; Using the phase delay information corresponding to each effective sound source direction in the directional delay parameter table to drive a multi-channel adaptive filter group, perform cascade phase compensation, and obtain a phase-corrected filter group output signal; The output signal of the filter group after the phase correction is subjected to inverse phase transformation and automatic gain control adjustment to generate a direction-dependent inverse phase signal group.

6. The active noise reduction method for wireless headphones according to claim 5, characterized in that: The method of using the phase delay information corresponding to each effective sound source direction in the direction delay parameter table to drive the multi-channel adaptive filter group, performing cascade phase compensation, and obtaining a phase-corrected filter group output signal includes: Extract the phase delay value corresponding to each effective sound source direction from the direction delay parameter table, and calculate the phase delay distribution of each effective sound source direction at different frequency points in combination with the frequency segmentation parameter; generating all-pass filter coefficients corresponding to each effective sound source direction based on the phase delay distribution, and constructing an all-pass filter bank matching the number of sound source directions; cascade-connecting the all-pass filter bank and the corresponding adaptive finite impulse response filters in the multi-channel adaptive filter bank according to a direction matching principle to form a cascade filter bank architecture; The cascaded filter bank architecture is driven to perform a joint process of amplitude filtering and phase correction on an input signal, and output a filter bank output signal after phase correction.

7. The active noise reduction method for a wireless headset according to claim 1, wherein: The method of using the intensity information in the sound source direction distribution data as a weighting coefficient to perform weighted synthesis on the directional inverted signal group to form a comprehensive inverted signal and applying a feedforward delay correction amount to output an adaptive noise reduction signal includes: Extracting intensity information corresponding to each effective sound source direction from the sound source direction distribution data, and calculating a weighting coefficient corresponding to each effective sound source direction through a normalization operation; Based on the angle difference between the directions of each effective sound source, the cosine function is used to calculate and construct the inter-directional interference suppression matrix. When the angle between adjacent directions is less than the preset threshold, the corresponding interference suppression factor is activated. Applying the weighting coefficient and the interference suppression factor to the direction-specific inverted signal group simultaneously to perform a weighted linear combination operation to generate an initial inverted signal; Applying a time domain window function smoothing process to the initial inverted signal and establishing a dynamic power monitoring mechanism to perform adaptive gain adjustment, and outputting a comprehensive inverted signal; The sound source direction change information in the continuous time window is monitored and a feedforward delay correction amount is applied to the integrated inverted signal to output an adaptive noise reduction signal.

8. The active noise reduction method for wireless headphones according to claim 7, wherein: The monitoring of the sound source direction change information within the continuous time window and applying a feedforward delay correction amount to the integrated inverted signal to output an adaptive noise reduction signal includes: Calculating the angular velocity value of each effective sound source based on the sound source direction change information of the sound source direction distribution data within the continuous time window, and generating the sound source position prediction data at the next moment through a first-order prediction filter operation; Inputting the sound source position prediction data into a Kalman filter to optimize the prediction accuracy to obtain a predicted angle change, and calculating a feedforward delay correction amount based on the predicted angle change and a partial derivative of the delay with respect to the angle; The direction change amplitude of the dominant sound source within adjacent sampling periods is monitored and a direction switching detection mechanism is established. When the direction change exceeds a preset angle threshold and the duration meets the switching condition, an exponential decay function is used to perform interpolation transition on the filter coefficients to obtain the interpolation transition processing result; Dynamically update the delay parameters of the integrated inverted signal according to the feedforward delay correction amount and the interpolation transition processing result, and output an adaptive noise reduction signal.

9. An active noise reduction device for a wireless headphone, characterized in that: The steps for implementing the active noise reduction method of the wireless headphone according to any one of claims 1 to 8, wherein the active noise reduction device of the wireless headphone comprises: A calculation module is used to perform time-division multiplexing sampling through the dual microphone arrays in the left and right acoustic cavities of the headset, obtain sampling data from the four-channel microphone array, and perform correlation function beamforming calculations to generate sound source direction distribution data; An establishment module is used to calculate the headphone geometric acoustic path difference using the direction information in the sound source direction distribution data, and establish a directional delay parameter table corresponding to each effective sound source direction; A directional filtering module is used to drive a multi-channel adaptive filter group according to the directional delay parameter table, perform directional filtering on the four-channel microphone array sampling data, and generate a directional anti-phase signal group; The output module is used to perform weighted synthesis on the directional inverted signal group using the intensity information in the sound source direction distribution data as a weighting coefficient to form a comprehensive inverted signal and apply a feedforward delay correction amount to output an adaptive noise reduction signal.

10. A wireless headset, characterized in that: The wireless headset is used to implement the steps of the active noise reduction method for a head-mounted wireless headset according to any one of claims 1 to 8.

Images