An adaptive array microphone noise reduction device

Abstract

The application provides a self-adaptive array microphone noise reduction device, and relates to the technical field of audio processing.The application is provided with a signal preprocessing module, a self-adaptive filter set, a DOA direction estimation module, a filter output selection module, a further noise reduction module and an OLA overlap addition module on the upper part of the array microphone; when the noise signal frequently changes in some positions, the multi-direction auxiliary filter is configured according to the statistical information obtained from the estimated DOA or the prior configuration.The auxiliary filter is always updated according to the DOA information in the background.In this way, the filter can quickly play a role in noise reduction without re-tracking from the beginning.

Description

Technical Field

[0001] This invention relates to the field of audio processing technology, and more particularly to an adaptive array microphone noise reduction device. Background Technology

[0002] A noise-canceling microphone is a type of microphone that uses directional processing of sound signals to improve the signal-to-noise ratio, enabling users to hear people speaking clearly even in the presence of background noise. Its basic principle is to assemble two microphones within the same hearing aid according to strict acoustic principles, allowing sound signals arriving from different angles to be amplified differently, thereby enhancing the useful signal and relatively reducing background noise.

[0003] In existing microphone array noise reduction schemes, common methods include Generalized Sidelobe Cancellation (GSC), minimum variance distortionless response (MVDR), or variations thereof. GSC is particularly prevalent in dual-microphone arrays. A standard GSC adaptive noise reduction scheme includes an Initialize Filter (FBF) and an ABM. The ABM adaptively eliminates noise, including both steady-state and non-steady-state noise, based on the output of the blocking matrix. An adaptive filter for the directional direction dynamically generates zeros for non-directional directions to achieve noise reduction.

[0004] However, existing microphone array noise reduction solutions suffer from poor dynamic noise reduction performance. For example, when the noise source is changing rapidly, the filter cannot be updated in time, and the filter coefficients cannot converge, resulting in poor noise reduction performance.

[0005] Therefore, it is necessary to provide an adaptive array microphone noise reduction device to solve the above-mentioned technical problems. Summary of the Invention

[0006] To address one of the aforementioned technical problems, this invention provides an adaptive array microphone noise reduction device, deployed on an array microphone. The array microphone includes several microphone acquisition units that acquire external audio signals to obtain array microphone signal data. The array microphone also includes a signal preprocessing module, an adaptive filter bank, a DOA direction estimation module, a filter output selection module, a further noise reduction module, and an OLA overlap and addition module.

[0007] Specifically, during adaptive array microphone noise reduction, the signal preprocessing module performs noise reduction preprocessing on the array microphone signal data to obtain noise reduction preprocessed data; the noise reduction preprocessed data is then used by the DOA direction estimation module to estimate the direction of arrival (DOA) to obtain direction estimation data; the adaptive filter bank includes a main adaptive filter and several auxiliary filters; the noise reduction preprocessed data and the direction estimation data are input together into the main adaptive filter and each auxiliary filter to obtain the output of the adaptive filter bank, and the filter coefficients of each filter in the adaptive filter bank are updated; the filter output selection module selects the signal from the output of the adaptive filter bank to obtain the optimal output signal; the further noise reduction module performs further noise reduction processing on the optimal output signal to obtain a further noise-reduced signal; the OLA overlap and addition module overlaps and adds the further noise-reduced signal and the optimal output signal to obtain the final output signal.

[0008] As a further solution, the signal preprocessing module performs frame segmentation, windowing, and FFT processing on the array microphone signal data;

[0009] Frame segmentation: Set the frame segmentation interval and perform frame segmentation operation on the array microphone signal according to the set frame segmentation interval to obtain several frame data;

[0010] Windowing: Hamming windows are used to window the framed data to prevent frequency leakage; the Hamming window function is as follows:

[0011]

[0012] Where n represents the truncated signal; α represents the Hamming window constant; and N-1 represents the truncated window length of the Hamming window.

[0013] FFT processing: Perform Fast Fourier Transform on the windowed frame data to obtain the corresponding noise reduction preprocessed data.

[0014] As a further solution, the DOA direction estimation module adopts the GCC-PHAT sound source localization method, which locates the sound source angle of the continuous sound source signals collected by the array microphones. The sound source signals are multiple sound sources that do not emit sound at the same time.

[0015] As a further solution, the adaptive filter bank performs filtering operations through the following steps:

[0016] Step A1: Set the main filtering direction of the main filter and the auxiliary filtering zero direction of each auxiliary filter;

[0017] Step A2: The main filter forms a gain peak in the main direction;

[0018] Step A3: The main filter adaptively forms valleys in other directions;

[0019] Step A4: Each auxiliary filter generates a gain peak in the corresponding auxiliary filter zero direction;

[0020] Step A5: Each auxiliary filter forms a valley in the corresponding auxiliary filter zero direction at 90° / 180°.

[0021] Step A6: Determine if there are more noise sources. If so, form a valley in the given direction.

[0022] As a further solution, the adaptive filter bank reinitializes each filter when it detects that the filter is diverging.

[0023] As a further solution, the adaptive filter bank determines whether to update the filter parameters based on the direction estimation data from the DOA direction estimation module:

[0024] Master filter update:

[0025] If the DOA direction estimation module estimates the direction of arrival α to be in or near the main filtering direction, then the noise estimation filter coefficients are updated and the target signal estimation filter coefficients are frozen; if the estimated direction of arrival α deviates from the main filtering direction by more than a threshold, then the noise estimation filter coefficients are frozen and the target signal estimation filter coefficients are updated.

[0026] Auxiliary filter update:

[0027] If the direction of arrival α estimated by the DOA estimation module is in or near the corresponding auxiliary filter zero direction, then the noise estimation filter coefficients are updated and the target signal estimation filter coefficients are frozen; if the estimated direction of arrival α deviates from the corresponding auxiliary filter zero direction by more than a threshold, then the noise estimation filter coefficients are frozen and the target signal estimation filter coefficients are updated.

[0028] As a further solution, the filter output selection module selects the output with the maximum signal energy and the minimum noise energy, and performs smoothing processing on the output signal.

[0029] As a further solution, the additional noise reduction module performs further noise reduction processing on the optimal output signal, including GSC adaptive noise reduction, minimum variance distortionless response, or other noise reduction modules.

[0030] As a further solution, the OLA overlap-add module divides the input signal data into fixed-length data blocks; each time an FFT is performed, the fixed-length data block is divided into several sub-blocks, and the latest input sub-block data is added to the FFT processed data sequence each time, while the last sub-block is discarded; each time an FFT is performed, there is an overlap of 3 sub-blocks; when outputting after FFT processing, only the data of the last processed sub-block is output each time.

[0031] As a further solution, the adaptive array microphone noise reduction device is deployed on a chip-level device and performs adaptive array microphone noise reduction on the array microphone.

[0032] Compared with related technologies, the adaptive array microphone noise reduction device provided by the present invention has the following beneficial effects:

[0033] Existing microphone array noise reduction solutions suffer from poor dynamic noise reduction performance, inability to update filters in a timely manner, and failure to converge filter coefficients. The main innovations of this embodiment include:

[0034] This invention deploys a signal preprocessing module, an adaptive filter bank, a DOA direction estimation module, a filter output selection module, a further noise reduction module, and an OLA overlap and summation module on an array microphone. When the noise signal frequently changes at certain locations, multi-directional auxiliary filters are configured based on pre-configured information or statistical information obtained from the estimated DOA. The auxiliary filters are continuously updated in the background based on the DOA information. This eliminates the need to re-track from scratch, allowing the filters to quickly take effect and reduce noise. Attached Figure Description

[0035] Figure 1 A flowchart illustrating the noise reduction process of an adaptive array microphone noise reduction device provided in an embodiment of the present invention;

[0036] Figure 2 A schematic diagram of an adaptive array microphone noise reduction device provided in an embodiment of the present invention. Figure 1 ;

[0037] Figure 3 A schematic diagram of an adaptive array microphone noise reduction device provided in an embodiment of the present invention. Figure 2 . Detailed Implementation

[0038] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0039] like Figure 1As shown in the figure, this embodiment provides an adaptive array microphone noise reduction device, which is deployed on an array microphone. The array microphone includes several microphone acquisition units, which acquire external audio signals through the array microphone to obtain array microphone signal data. The array microphone is also equipped with a signal preprocessing module, an adaptive filter bank, a DOA direction estimation module, a filter output selection module, a further noise reduction module, and an OLA overlap and addition module.

[0040] Specifically, during adaptive array microphone noise reduction, the signal preprocessing module performs noise reduction preprocessing on the array microphone signal data to obtain noise reduction preprocessed data; the noise reduction preprocessed data is then used by the DOA direction estimation module to estimate the direction of arrival (DOA) to obtain direction estimation data; the adaptive filter bank includes a main adaptive filter and several auxiliary filters; the noise reduction preprocessed data and the direction estimation data are input together into the main adaptive filter and each auxiliary filter to obtain the output of the adaptive filter bank, and the filter coefficients of each filter in the adaptive filter bank are updated; the filter output selection module selects the signal from the output of the adaptive filter bank to obtain the optimal output signal; the further noise reduction module performs further noise reduction processing on the optimal output signal to obtain a further noise-reduced signal; the OLA overlap and addition module overlaps and adds the further noise-reduced signal and the optimal output signal to obtain the final output signal.

[0041] It should be noted that in this invention, a primary adaptive filter and several auxiliary adaptive filters are defined. This invention utilizes multiple auxiliary filters to simultaneously track multiple attenuation directions, thereby achieving rapid convergence.

[0042] After receiving the signal, the array undergoes necessary data processing, including framing, windowing, and FFT, before first performing direction-of-arrival (DOA) estimation via the DOA module. This module also controls the update process of the main filter and each auxiliary filter.

[0043] As a further solution, the signal preprocessing module performs frame segmentation, windowing, and FFT processing on the array microphone signal data;

[0044] Frame segmentation: Set the frame segmentation interval and perform frame segmentation operation on the array microphone signal according to the set frame segmentation interval to obtain several frame data;

[0045] Windowing: Hamming windows are used to window the framed data to prevent frequency leakage; the Hamming window function is as follows:

[0046]

[0047] Where n represents the truncated signal; α represents the Hamming window constant; and N-1 represents the truncated window length of the Hamming window.

[0048] FFT processing: Perform Fast Fourier Transform on the windowed frame data to obtain the corresponding noise reduction preprocessed data.

[0049] As a further solution, the DOA direction estimation module adopts the GCC-PHAT sound source localization method, which locates the sound source angle of the continuous sound source signals collected by the array microphones. The sound source signals are multiple sound sources that do not emit sound at the same time.

[0050] It should be noted that the estimation method used is GCC-PHAT, which is a commonly used algorithm and will not be explained in detail here.

[0051] As a further solution, the adaptive filter bank performs filtering operations through the following steps:

[0052] Step A1: Set the main filtering direction of the main filter and the auxiliary filtering zero direction of each auxiliary filter;

[0053] Step A2: The main filter forms a gain peak in the main direction;

[0054] Step A3: The main filter adaptively forms valleys in other directions;

[0055] Step A4: Each auxiliary filter generates a gain peak in the corresponding auxiliary filter zero direction;

[0056] Step A5: Each auxiliary filter forms a valley in the corresponding auxiliary filter zero direction at 90° / 180°.

[0057] Step A6: Determine if there are more noise sources. If so, form a valley in the given direction.

[0058] It should be noted that: such as Figure 2 and Figure 3 As shown, the update criteria are as follows: taking the 0° (main filter direction) direction as an example, other directions are similar. If the estimated direction of arrival α is near 0°, the noise estimation filter coefficients are updated, and the target signal estimation filter coefficients are frozen; if the estimated direction of arrival α deviates from 0° by more than a threshold (e.g., ±30°), the noise estimation filter coefficients are frozen, and the target signal estimation filter coefficients are updated. This rule is applied to both the main filter and the auxiliary filter bank.

[0059] Figure 2 This indicates a dual-microphone (or multi-microphone) headset, with one main filter and two auxiliary filters. The main filter is for the direction of the mouth, the zero point of auxiliary filter 1 is for the direction of the back of the head, and the zero point of auxiliary filter 2 is for the direction of the side.

[0060] Figure 3This indicates a dual-microphone system with one main filter and two auxiliary filters. The main filter is directed towards the mouth, the zero point of auxiliary filter 1 is directed towards the speaker, and the zero point of auxiliary filter 2 is perpendicular to the line connecting the mouth and the speaker.

[0061] As a further solution, the adaptive filter bank reinitializes each filter when it detects that the filter is diverging.

[0062] As a further solution, the adaptive filter bank determines whether to update the filter parameters based on the direction estimation data from the DOA direction estimation module:

[0063] Master filter update:

[0064] If the DOA direction estimation module estimates the direction of arrival α to be in or near the main filtering direction, then the noise estimation filter coefficients are updated and the target signal estimation filter coefficients are frozen; if the estimated direction of arrival α deviates from the main filtering direction by more than a threshold, then the noise estimation filter coefficients are frozen and the target signal estimation filter coefficients are updated.

[0065] Auxiliary filter update:

[0066] If the direction of arrival α estimated by the DOA estimation module is in or near the corresponding auxiliary filter zero direction, then the noise estimation filter coefficients are updated and the target signal estimation filter coefficients are frozen; if the estimated direction of arrival α deviates from the corresponding auxiliary filter zero direction by more than a threshold, then the noise estimation filter coefficients are frozen and the target signal estimation filter coefficients are updated.

[0067] As a further solution, the filter output selection module selects the output with the maximum signal energy and the minimum noise energy, and performs smoothing processing on the output signal.

[0068] As a further solution, the additional noise reduction module performs further noise reduction processing on the optimal output signal, including GSC adaptive noise reduction, minimum variance distortionless response, or other noise reduction modules.

[0069] As a further solution, the OLA overlap-add module divides the input signal data into fixed-length data blocks; each time an FFT is performed, the fixed-length data block is divided into several sub-blocks, and the latest input sub-block data is added to the FFT processed data sequence each time, while the last sub-block is discarded; each time an FFT is performed, there is an overlap of 3 sub-blocks; when outputting after FFT processing, only the data of the last processed sub-block is output each time.

[0070] As a further solution, the adaptive array microphone noise reduction device is deployed on a chip-level device and performs adaptive array microphone noise reduction on the array microphone.

[0071] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. An adaptive array microphone noise reduction device, characterized in that, Deployed on an array microphone, the array microphone includes several microphone acquisition units that acquire external audio signals to obtain array microphone signal data; the array microphone also deploys a signal preprocessing module, an adaptive filter bank, a DOA direction estimation module, a filter output selection module, a further noise reduction module, and an OLA overlap and addition module. When performing adaptive array microphone noise reduction, the signal preprocessing module performs noise reduction preprocessing on the array microphone signal data to obtain noise reduction preprocessed data. The noise reduction preprocessing data is used to estimate the direction of arrival (DOA) through the DOA estimation module to obtain direction estimation data. The adaptive filter bank includes a main adaptive filter and several auxiliary filters. The noise reduction preprocessing data and the direction estimation data are input together into the main adaptive filter and each auxiliary filter to obtain the output of the adaptive filter bank, and the filter coefficients of each filter in the adaptive filter bank are updated. The filter output selection module selects the signal from the output of the adaptive filter bank to obtain the optimal output signal. The further noise reduction module performs further noise reduction processing on the optimal output signal to obtain a further noise-reduced signal. The OLA overlap and addition module overlaps and adds the further noise-reduced signal and the optimal output signal to obtain the final output signal. The adaptive filter bank performs filtering operations through the following steps: Step A1: Set the main filtering direction of the main filter and the auxiliary filtering zero direction of each auxiliary filter; Step A2: The main filter forms a gain peak in the main direction; Step A3: The main filter adaptively forms valleys in other directions; Step A4: Each auxiliary filter generates a gain peak in the corresponding auxiliary filter zero direction; Step A5: Each auxiliary filter forms a valley in the corresponding auxiliary filter zero direction at 90° / 180°. Step A6: Determine if there are more noise sources. If so, form a valley in the given direction. The adaptive filter bank determines whether to update the filter parameters based on the direction estimation data from the DOA direction estimation module. Master filter update: If the direction of arrival α estimated by the DOA estimation module is in or near the main filtering direction, then the noise estimation filter coefficients are updated and the target signal estimation filter coefficients are frozen; if the estimated direction of arrival α deviates from the main filtering direction by more than a threshold, then the noise estimation filter coefficients are frozen and the target signal estimation filter coefficients are updated. Auxiliary filter update: If the direction of arrival α estimated by the DOA estimation module is in or near the corresponding auxiliary filter zero direction, then the noise estimation filter coefficients are updated and the target signal estimation filter coefficients are frozen; if the estimated direction of arrival α deviates from the corresponding auxiliary filter zero direction by more than a threshold, then the noise estimation filter coefficients are frozen and the target signal estimation filter coefficients are updated. The OLA overlap-add module divides the input signal data into fixed-length data blocks. Each time an FFT is performed, the fixed-length data block is divided into several sub-blocks. Each time, the latest input sub-block data is added to the data sequence of the FFT process, and the last sub-block is discarded. Each time an FFT process is performed, there is an overlap of 3 sub-blocks. When outputting after FFT processing, only the data of the last sub-block after processing is output each time.

2. The adaptive array microphone noise reduction device according to claim 1, characterized in that, The signal preprocessing module performs frame segmentation, windowing, and FFT processing on the array microphone signal data. Frame segmentation: Set the frame segmentation interval and perform frame segmentation operation on the array microphone signal according to the set frame segmentation interval to obtain several frame data; Windowing: Hamming windows are used to window the framed data to prevent frequency leakage; the Hamming window function is as follows: Where n represents the truncated signal; α represents the Hamming window constant; and N-1 represents the truncated window length of the Hamming window. FFT processing: Perform Fast Fourier Transform on the windowed frame data to obtain the corresponding noise reduction preprocessed data.

3. The adaptive array microphone noise reduction device according to claim 1, characterized in that, The DOA direction estimation module uses the GCC-PHAT sound source localization method, which locates the sound source angle of the continuous sound source signals collected by the array microphones. The sound source signals are multiple sound sources that do not emit sound at the same time.

4. The adaptive array microphone noise reduction device according to claim 1, characterized in that, When the adaptive filter bank detects that the filter is diverging, it reinitializes each filter.

5. The adaptive array microphone noise reduction device according to claim 1, characterized in that, The filter output selection module selects the output with the highest signal energy and the lowest noise energy, and performs smoothing processing on the output signal.

6. The adaptive array microphone noise reduction device according to claim 1, characterized in that, The further noise reduction module performs further noise reduction processing on the optimal output signal, including GSC adaptive noise reduction, minimum variance distortionless response, or other noise reduction modules.

7. The adaptive array microphone noise reduction device according to claim 1, characterized in that, The adaptive array microphone noise reduction device is deployed on a chip-level device and performs adaptive array microphone noise reduction on the array microphone.

Images