A directional microphone and method for reducing echo failure for video conferencing

Abstract

The application discloses a directional microphone and a method for reducing echo failure in a video conference, and belongs to the technical field of audio processing. The microphone comprises a microphone rod and a microphone head, and the periphery of the microphone head is sequentially provided with an internal sound insulation layer, a protective cover and an external sound insulation layer. The protective cover is divided into first to third sound resistance zones in the axial direction, and the porosity of each sound resistance zone increases gradually; the density of the internal and external sound insulation layers is inversely correlated with the porosity of the corresponding sound resistance zone and is distributed in a gradient manner. The application further discloses an echo cancellation method: a gradient sound resistance transfer function compensation structure is constructed to compensate for structural distortion, a dynamic time delay is solved, the transfer function is embedded into an adaptive filter as a prior acoustic constraint to perform echo estimation, and a threshold is dynamically configured based on the porosity ratio to suppress residual echo. The application is deeply coupled with acoustic structure and signal processing, and can significantly improve the directional sound pickup accuracy and echo suppression effect, thereby guaranteeing the call quality of the video conference.

Description

Technical Field

[0001] This invention relates to the field of audio processing technology, and more particularly to a directional microphone for video conferencing and a method for reducing echo faults. Background Technology

[0002] In video conferencing systems, audio quality directly impacts communication efficiency and user experience. Current video conferencing equipment generally faces two major technical challenges: first, microphones struggle to achieve precise directional pickup in complex meeting environments, easily picking up ambient noise and interfering speech from non-target directions; second, distant speech played through speakers is re-picked up by nearby microphones, creating echoes that cause audio trailing, feedback, or even prevent the meeting from proceeding normally.

[0003] To improve directivity, existing technologies typically employ microphone arrays combined with beamforming algorithms. However, array design is costly, algorithms are complex, and deployment in miniaturized conferencing equipment is limited. Some solutions attempt to add pop filters or windscreens made of a single material to the front of the microphone, but such structures only provide simple physical protection and lack the ability to finely control the sound wave transmission path, making it difficult to effectively suppress lateral and rearward interference.

[0004] Traditional methods for addressing echo problems rely on adaptive filters (such as the NLMS algorithm) to model the echo path. However, in real-world meeting scenarios, the acoustic path between the speaker and microphone exhibits time-varying characteristics and is affected by factors such as the device's casing structure and the angle of sound wave incidence. This results in slow filter convergence and large steady-state errors, especially during two-way calls where filter divergence can easily occur, leading to speech distortion or echo remnants. In existing technologies, signal processing algorithms and microphone physical structures are often designed independently, failing to form an effective collaborative mechanism, making it difficult to overcome the bottleneck in echo suppression performance.

[0005] In summary, designing a microphone system that combines strong directivity with efficient echo suppression from the perspective of integrating acoustic structure and signal processing is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] Purpose of the invention:

[0007] This invention aims to provide a directional microphone for video conferencing and a method for reducing echo faults. By deeply coupling acoustic gradient structures and signal processing algorithms, it achieves precise directional sound pickup and efficient echo suppression, thereby improving the audio quality of video conferencing.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] A directional microphone for video conferencing includes a microphone boom and a microphone head, wherein an inner sound insulation layer, a protective cover, and an outer sound insulation layer are arranged sequentially from the inside to the outside around the microphone head;

[0010] The protective cover is divided into a first acoustic resistance zone, a second acoustic resistance zone, and a third acoustic resistance zone along the axial direction from the near end to the far end. Each acoustic resistance zone has honeycomb holes on its surface, and the porosity increases from the first acoustic resistance zone to the third acoustic resistance zone. The first acoustic resistance zone has the lowest porosity, and the third acoustic resistance zone has the highest porosity.

[0011] The density of the inner and outer sound insulation layers is inversely correlated with the porosity of the corresponding acoustic resistance zones, meaning that the density of the sound insulation layer corresponding to the third acoustic resistance zone is the lowest, and the density of the sound insulation layer corresponding to the first acoustic resistance zone is the highest.

[0012] Furthermore, a vibration isolation ring is provided between the microphone rod and the protective cover.

[0013] Furthermore, the honeycomb porosity ratio of the third acoustic resistance zone, the second acoustic resistance zone, and the first acoustic resistance zone is 1:1, and the axial length of each acoustic resistance zone accounts for 20%-40%, 30%-50%, and 20%-40% of the total length of the protective cover, respectively.

[0014] Furthermore, the honeycomb aperture of the protective cover gradually decreases from the third acoustic resistance zone to the first acoustic resistance zone, and the honeycomb aperture in the third acoustic resistance zone is a through-hole structure, while the honeycomb aperture in the first acoustic resistance zone is a blind hole or micro-pore structure.

[0015] A method for reducing echo failures in video conferencing includes the following steps:

[0016] S1: Obtain the porosity of each of the first to third acoustic resistance zones in the directional microphone shield. With sound insulation layer density ,in Based on the porosity and density parameters, a frequency-related parameter is constructed. and the angle of incidence of sound waves Related gradient acoustic drag transfer function ;

[0017] S2: Real-time acquisition of far-end speech reference signal output from local speaker. and mixed signals picked up by near-end directional microphones In the frequency domain, the gradient acoustic impedance transfer function With mixed signals Perform convolution operations to compensate for the fixed acoustic distortion caused by the physical structure of the protective shield, and obtain the compensated signal. The far-end speech reference signal is calculated based on the cross-correlation function. With the compensated signal Dynamic delay parameters between ;

[0018] S3: Based on the dynamic delay parameters For remote voice reference signal Perform time delay alignment to generate a time delay aligned reference signal. Construct a coupled adaptive filter and apply the gradient acoustic impedance transfer function. The echo estimation signal is generated by embedding the prior acoustic constraints into the update rules of the filter coefficients. ;

[0019] S4: The compensated signal Subtract the echo estimation signal The error signal is obtained. The residual echo suppression threshold is dynamically configured based on the porosity ratio of each acoustic impedance zone, and the error signal is then processed based on this threshold. Nonlinear residual echo suppression processing is performed to ultimately output a clean near-end speech signal.

[0020] Furthermore, in S1, the porosity of the three acoustic resistance zones of the microphone and the corresponding density parameters of the sound insulation layer are first measured and collected. The two types of parameters are normalized to eliminate the differences in the dimensions and numerical magnitudes of the parameters, resulting in standardized normalized porosity and normalized sound insulation layer density. Then, based on the theory of sound wave propagation, combined with the characteristics of porosity gradient distribution and density inverse correlation distribution, parameters such as acoustic resistance zone weight coefficient, sound velocity in air, axial length of acoustic resistance zone, imaginary unit, and cosine value of sound wave incident angle are incorporated to construct a frequency domain-angle domain joint gradient acoustic resistance transfer function that takes into account both sound wave frequency and incident angle. This is used to quantify the filtering and attenuation effect of the protective cover structure on sound waves.

[0021] Furthermore, in S2, a fast Fourier transform is first performed on the near-end mixed signal to convert the time-domain signal into a frequency-domain signal. Then, a frequency-domain convolution operation is performed with the gradient acoustic impedance transfer function to cancel the inherent distortion of the structure. Subsequently, an inverse fast Fourier transform is performed on the frequency-domain compensated signal to restore the time-domain compensated signal. Next, a cross-correlation operation is performed on the far-end reference signal and the time-domain compensated signal. By calculating the time delay value corresponding to the maximum value of the cross-correlation function, the index corresponding to the cross-correlation peak value is extracted to determine the dynamic time delay parameter in units of sampling points.

[0022] Furthermore, in S3, the far-end reference signal is first shifted and aligned based on the obtained dynamic time delay parameters to obtain an aligned reference signal that matches the time sequence. Then, based on the normalized least mean square algorithm, the gradient acoustic impedance transfer function is used as a priori acoustic constraint and embedded in the update rule of the filter weights. Combined with the step size factor, regularization constant, and the L2 squared correlation parameter of the aligned reference signal, the filter weights are iteratively updated. Finally, the echo estimation signal is output by performing a convolution operation between the transpose of the filter weight vector and the aligned reference signal.

[0023] Furthermore, in S4, the echo component subtraction operation is first performed, and the difference between the time-domain compensated signal and the echo estimation signal is calculated to obtain the error signal that initially eliminates the main echo. Then, based on the porosity ratio of the three acoustic resistance zones of the microphone, the residual echo suppression threshold is dynamically configured in combination with the threshold adjustment coefficient. Then, nonlinear amplitude determination processing is performed on the error signal. If the amplitude of the error signal is less than or equal to the dynamic threshold, it is set to zero. If the amplitude is greater than the threshold, the original signal value is retained. Finally, the processed near-end speech signal is output.

[0024] Furthermore, the gradient acoustic impedance transfer function :

[0025]

[0026] in:

[0027] The acoustic resistance zone weighting coefficient is positively correlated with the proportion of the axial length of each acoustic resistance zone, satisfying the following condition: ;

[0028] Speed ​​of sound in air;

[0029] : No. Axial length of each acoustic resistance zone;

[0030] Imaginary unit, ;

[0031] : The cosine of the incident angle of the sound wave, which characterizes the attenuation gain of the obliquely incident sound wave.

[0032] Working principle:

[0033] The directional microphone achieves directional sound pickup through an axial gradient acoustic impedance structure and an inversely correlated density sound insulation layer. The protective cover is divided into three acoustic impedance zones along the axis, with porosity gradually increasing from the first zone near the end to the third zone far the end. This, combined with a differentiated honeycomb structure, forms a gradient sound wave transmission channel. The density of the inner and outer sound insulation layers and the porosity of the corresponding acoustic impedance zones exhibit an inversely correlated gradient distribution, effectively blocking conference echoes and environmental noise from directions other than the main axis, ensuring only the main axis voice is smoothly transmitted to the microphone. Simultaneously, a vibration isolation ring between the microphone boom and the protective cover blocks mechanical vibration transmission, preventing mechanical noise from interfering with the pickup signal. This ensures directional pickup of pure voice and suppression of echo interference throughout the entire process.

[0034] The present invention has the following beneficial effects:

[0035] 1. This invention features a protective cover with a porosity gradient distribution around the microphone head, combined with a gradient distribution of inner and outer sound insulation layers whose density and porosity are inversely correlated, forming an axial sound wave transmission channel and a non-axial sound wave attenuation barrier. When sound waves are incident along the main axis, they pass through the high-porosity acoustic resistance zone, resulting in high transmission efficiency. Meanwhile, sound waves traveling in the non-main axis direction undergo multiple reflections and energy attenuation when passing through the gradient structure with alternating porosity and density, thus achieving excellent directional sound pickup and effectively suppressing lateral noise and interference from far-end speaker leakage in the conference environment.

[0036] 2. This invention quantifies the gradient acoustic impedance characteristics of the microphone shield into a gradient acoustic impedance transfer function, and embeds it as a priori acoustic constraint into the coefficient update rule of the adaptive filter. Compared with traditional filtering algorithms that rely solely on statistical information, this scheme ensures that the filter always follows physical acoustic laws during the iteration process, significantly improving the accuracy and convergence speed of echo path modeling. Especially in complex scenarios such as two-way calls, it effectively avoids filter divergence problems, ensuring the stability of echo estimation and distortion-free transmission of near-end speech.

[0037] 3. This invention, based on traditional adaptive filtering echo cancellation, further introduces a nonlinear residual echo suppression threshold based on the dynamic configuration of the porosity ratio of the microphone's three acoustic resistance zones. This threshold can adaptively adjust according to the microphone's own acoustic characteristics, effectively eliminating residual echoes in strong echo scenarios while avoiding excessive suppression during near-end speech activity. This achieves a good balance between suppressing residual echoes and maintaining the naturalness of speech, outputting a clean, clear, and tailless near-end speech signal. Attached Figure Description

[0038] Figure 1 This is an internal structural diagram of a directional microphone for video conferencing proposed in this invention.

[0039] Figure 2 This is a structural diagram of a protective cover for a directional microphone used in video conferencing, as proposed in this invention.

[0040] Figure 3 This is a flowchart of a method for reducing echo failures in video conferencing proposed in this invention.

[0041] Legend:

[0042] 1. Microphone boom; 2. Microphone head; 3. Internal sound insulation layer; 4. Protective cover; 401. First acoustic resistance zone; 402. Second acoustic resistance zone; 403. Third acoustic resistance zone; 5. External sound insulation layer; 6. Vibration isolation ring. Detailed Implementation

[0043] The present invention will be further described in detail below with reference to specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Example 1

[0045] A directional microphone for video conferencing includes a microphone boom 1 and a microphone head 2. The microphone head 2 is provided with an inner sound insulation layer 3, a protective cover 4 and an outer sound insulation layer 5 arranged sequentially from the inside to the outside.

[0046] The protective cover 4 is divided into a first acoustic resistance zone 401, a second acoustic resistance zone 402 and a third acoustic resistance zone 403 along the axial direction from the near end to the far end. Each acoustic resistance zone has honeycomb holes on its surface, and the porosity increases from the first acoustic resistance zone 401 to the third acoustic resistance zone 403. The first acoustic resistance zone 401 has the lowest porosity and the third acoustic resistance zone 403 has the highest porosity.

[0047] The density of the inner sound insulation layer 3 and the outer sound insulation layer 5 are inversely correlated with the porosity of the corresponding acoustic resistance zone. That is, the density of the sound insulation layer corresponding to the third acoustic resistance zone 403 is the lowest, and the density of the sound insulation layer corresponding to the first acoustic resistance zone 401 is the highest.

[0048] A vibration isolation ring 6 is installed between the microphone pole 1 and the protective cover 4.

[0049] The honeycomb porosity ratio of the third acoustic resistance zone 403, the second acoustic resistance zone 402, and the first acoustic resistance zone 401 is 3-5:1.5-2.5:1, and the axial length of each acoustic resistance zone accounts for 20%-40%, 30%-50%, and 20%-40% of the total length of the protective cover 4, respectively.

[0050] The honeycomb aperture of the protective cover 4 gradually decreases from the third acoustic resistance zone 403 to the first acoustic resistance zone 401, and the honeycomb aperture of the third acoustic resistance zone 403 is a through-hole structure, while the honeycomb aperture of the first acoustic resistance zone 401 is a blind hole or micro-hole structure.

[0051] Example 2

[0052] A method for reducing echo failures in video conferencing includes the following steps:

[0053] S1: Obtain the porosity of each of the first to third acoustic resistance zones in the directional microphone shield. With sound insulation layer density ,in Based on the porosity and density parameters, a frequency-related parameter is constructed. and the angle of incidence of sound waves Related gradient acoustic drag transfer function ;

[0054] S2: Real-time acquisition of far-end speech reference signal output from local speaker. and mixed signals picked up by near-end directional microphones In the frequency domain, the gradient acoustic impedance transfer function With mixed signals Perform convolution operations to compensate for the fixed acoustic distortion caused by the physical structure of the protective shield, and obtain the compensated signal. The far-end speech reference signal is calculated based on the cross-correlation function. With the compensated signal Dynamic delay parameters between ;

[0055] S3: Based on the dynamic delay parameters For remote voice reference signal Perform time delay alignment to generate a time delay aligned reference signal. Construct a coupled adaptive filter and apply the gradient acoustic impedance transfer function. The echo estimation signal is generated by embedding the prior acoustic constraints into the update rules of the filter coefficients. ;

[0056] S4: The compensated signal Subtract the echo estimation signal The error signal is obtained. The residual echo suppression threshold is dynamically configured based on the porosity ratio of each acoustic impedance zone, and the error signal is then processed based on this threshold. Nonlinear residual echo suppression processing is performed to ultimately output a clean near-end speech signal.

[0057] In S1, the porosity of the three acoustic resistance zones of the microphone and the corresponding density parameters of the sound insulation layer are first measured and collected. The two types of parameters are normalized to eliminate the differences in the dimensions and numerical magnitudes of the parameters, resulting in standardized normalized porosity and normalized sound insulation layer density. Then, based on the sound wave propagation theory, combined with the porosity gradient distribution and density inverse correlation distribution characteristics, parameters such as acoustic resistance zone weighting coefficient, sound velocity in air, axial length of the acoustic resistance zone, imaginary unit, and cosine value of the sound wave incident angle are incorporated to construct a frequency domain-angle domain joint gradient acoustic resistance transfer function that takes into account both the sound wave frequency and the incident angle. This is used to quantify the filtering and attenuation effect of the protective cover structure on sound waves.

[0058] In S2, a fast Fourier transform is first performed on the near-end mixed signal to convert the time-domain signal into a frequency-domain signal. Then, a frequency-domain convolution operation is performed with the gradient acoustic impedance transfer function to cancel the inherent distortion of the structure. Subsequently, an inverse fast Fourier transform is performed on the frequency-domain compensated signal to restore the time-domain compensated signal. Next, a cross-correlation operation is performed on the far-end reference signal and the time-domain compensated signal. By calculating the time delay value corresponding to the maximum value of the cross-correlation function, the index corresponding to the cross-correlation peak value is extracted to determine the dynamic time delay parameter in units of sampling points.

[0059] In S3, the far-end reference signal is first shifted and aligned based on the obtained dynamic time delay parameters to obtain an aligned reference signal that matches the time sequence. Then, based on the normalized least mean square algorithm, the gradient acoustic impedance transfer function is used as a priori acoustic constraint and embedded in the update rule of the filter weights. Combined with the step size factor, regularization constant, and the L2 squared correlation parameter of the aligned reference signal, the filter weights are iteratively updated. Finally, the echo estimation signal is output by performing a convolution operation between the transpose of the filter weight vector and the aligned reference signal.

[0060] In S4, the echo component subtraction operation is first performed, and the difference between the time-domain compensated signal and the echo estimation signal is calculated to obtain the error signal for preliminary removal of the main echo. Then, based on the porosity ratio of the three acoustic resistance zones of the microphone, the residual echo suppression threshold is dynamically configured in combination with the threshold adjustment coefficient. Then, nonlinear amplitude determination processing is performed on the error signal. If the amplitude of the error signal is less than or equal to the dynamic threshold, it is set to zero. If the amplitude is greater than the threshold, the original signal value is retained. Finally, the processed near-end speech signal is output.

[0061] The gradient acoustic resistance transfer function :

[0062]

[0063] in:

[0064] The acoustic resistance zone weighting coefficient is positively correlated with the proportion of the axial length of each acoustic resistance zone, satisfying the following condition: ;

[0065] Speed ​​of sound in air;

[0066] : No. Axial length of each acoustic resistance zone;

[0067] Imaginary unit, ;

[0068] : The cosine of the incident angle of the sound wave, which characterizes the attenuation gain of the obliquely incident sound wave.

[0069] Example 3

[0070] A method for reducing echo failures in video conferencing includes the following steps:

[0071] S1: Constructing the gradient acoustic drag transfer function

[0072] By combining the spatial distribution characteristics of microphone gradient acoustic impedance and sound insulation layer density, an acoustic transfer function that takes into account both frequency and incident angle is established. The filtering and attenuation effects of the protective cover structure on sound waves are quantified, providing a priori basis for subsequent distortion compensation and filtering constraints.

[0073] 1. Parameter Acquisition and Normalization

[0074] The porosity of the microphone in three acoustic impedance zones was measured. Corresponding sound insulation layer density Porosity and density are normalized:

[0075]

[0076] In the formula: To normalize porosity, To normalize the density of the sound insulation layer, Corresponding to the first acoustic resistance zone, Corresponding to the second acoustic resistance zone, This corresponds to the third acoustic resistance zone.

[0077] 2. Gradient acoustic impedance transfer function formula

[0078] Based on sound wave propagation theory and combined with porosity gradient and density inverse correlation distribution, a joint frequency-angle domain gradient acoustic impedance transfer function is constructed. :

[0079]

[0080] in:

[0081] The acoustic resistance zone weighting coefficient is positively correlated with the proportion of the axial length of each acoustic resistance zone, satisfying the following condition: ;

[0082] Speed ​​of sound in air (taken as 340 m / s at room temperature);

[0083] : No. Axial length of each acoustic resistance zone;

[0084] Imaginary unit, ;

[0085] : The cosine of the incident angle of the sound wave, which characterizes the attenuation gain of the obliquely incident sound wave.

[0086] The physical meaning of the transfer function: It quantifies the amplitude attenuation and phase shift characteristics of the microphone gradient structure for sound waves of different frequencies and incident angles, and is the core basis for eliminating structural distortion.

[0087] S2: Compensate for fixed acoustic distortion and solve for dynamic time delay parameters

[0088] To address the speech signal distortion and echo propagation delay deviation caused by the physical structure of the microphone shield, frequency domain distortion compensation is performed first, and then the echo delay is accurately located through a cross-correlation algorithm, laying the foundation for subsequent filtering and alignment.

[0089] 1. Frequency domain convolution to compensate for acoustic distortion

[0090] For near-end mixed signals Perform a Fast Fourier Transform (FFT) to convert to the frequency domain. , with gradient acoustic drag transfer function Perform convolution operations to cancel out inherent structural distortions:

[0091]

[0092] Frequency domain compensation signal Perform an inverse fast Fourier transform (IFFT) to obtain the time-domain compensated signal:

[0093]

[0094] In the formula: For frequency domain convolution operators, To compensate for the frequency domain signal.

[0095] 2. Solving for dynamic time delay parameters using cross-correlation

[0096] Remote reference signal With the compensated signal Perform cross-correlation calculations; the index corresponding to the cross-correlation peak is the dynamic delay parameter. :

[0097]

[0098]

[0099] In the formula:

[0100] : Signal frame length (number of sampling points, usually 256 / 512 / 1024);

[0101] Cross-correlation function, which characterizes the similarity between two signals;

[0102] : Take the time delay value corresponding to the maximum value of the cross-correlation function ;

[0103] The unit is the sampling point, which reflects the dynamic time delay of the echo propagating from the speaker to the microphone.

[0104] S3: Gradient Coupling Adaptive Filtering

[0105] By embedding the gradient acoustic impedance transfer function as a prior constraint into the adaptive filter, the slow convergence and false convergence problems of traditional filtering are avoided, the echo component is accurately estimated, and the near-end speech is distortion-free.

[0106] 1. Reference signal delay alignment

[0107] Based on dynamic delay parameters For remote reference signal Perform shift alignment to eliminate propagation delay bias:

[0108]

[0109] In the formula: It serves as a time-delay aligned reference signal, perfectly synchronized with the echo signal timing.

[0110] 2. Construction of Coupled Adaptive Filters

[0111] Based on the Normalized Least Mean Square (NLMS) algorithm, a gradient acoustic drag transfer function is embedded. As a constraint for coefficient updates, the filter weights Update rules:

[0112]

[0113] The echo estimation signal is obtained by convolving the filter weights with the aligned reference signal:

[0114]

[0115] Explanation of parameters in the formula:

[0116] Step size factor (value 0 < <1, control the convergence speed of the filter);

[0117] Regularization constant (to avoid denominators being 0, with values ​​ranging from 1e-6 to 1e-8);

[0118] The square of the L2 norm of the reference signal is used for normalization.

[0119] : Transpose of the filter weight vector;

[0120] : Prior acoustic constraints ensure that the filter coefficients fit the acoustic characteristics of the microphone and suppress noise and false echoes in non-major axis directions.

[0121] S4: Echo Cancellation and Clean Voice Output

[0122] By subtracting the echo component and then dynamically configuring the suppression threshold based on the microphone porosity gradient, residual echoes are eliminated, and finally interference-free near-end speech is output.

[0123] 1. Echo component subtraction elimination

[0124] Compensated signal Subtract echo estimation signal The error signal without the main echo is obtained:

[0125]

[0126] 2. Dynamic threshold residual echo suppression

[0127] Based on the porosity ratio of the microphone's three acoustic resistance zones, the residual echo suppression threshold is dynamically configured. Threshold formula:

[0128]

[0129] For error signals Perform nonlinear suppression processing: when When it is determined to be a residual echo, it is set to zero; when When the signal is identified as near-end speech, its original value is preserved, and a clean signal is ultimately output.

[0130]

[0131] In the formula:

[0132] Threshold adjustment coefficient (valued from 0.5 to 1.2 depending on the ambient noise level of the meeting).

[0133] The adaptive residual echo threshold is adjusted in real time according to the microphone porosity gradient, balancing suppression effect and speech integrity.

[0134] The final output is a clean near-end voice signal with no echo and low noise, suitable for high-definition video conferencing calls.

[0135] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A directional microphone for video conferencing, comprising a microphone boom (1) and a microphone head (2), characterized in that: The microphone (2) is provided with an inner sound insulation layer (3), a protective cover (4) and an outer sound insulation layer (5) from the inside to the outside. The protective cover (4) is divided into a first acoustic resistance zone (401), a second acoustic resistance zone (402) and a third acoustic resistance zone (403) along the axial direction from the near end to the far end. Each acoustic resistance zone has honeycomb holes on its surface, and the porosity increases from the first acoustic resistance zone (401) to the third acoustic resistance zone (403). The first acoustic resistance zone (401) has the lowest porosity, and the third acoustic resistance zone (403) has the highest porosity. The density of the inner sound insulation layer (3) and the outer sound insulation layer (5) are inversely correlated with the porosity of the corresponding acoustic resistance zone. That is, the density of the sound insulation layer corresponding to the third acoustic resistance zone (403) is the lowest, and the density of the sound insulation layer corresponding to the first acoustic resistance zone (401) is the highest.

2. A directional microphone for video conferencing according to claim 1, characterized in that: A vibration isolation ring (6) is provided between the microphone rod (1) and the protective cover (4).

3. A directional microphone for video conferencing according to claim 1, characterized in that: The honeycomb porosity ratio of the third acoustic barrier zone (403), the second acoustic barrier zone (402), and the first acoustic barrier zone (401) is (3-5):(1.5-2.5):1, and the axial length of each acoustic barrier zone accounts for 20%-40%, 30%-50%, and 20%-40% of the total length of the protective cover (4), respectively.

4. A directional microphone for video conferencing according to claim 1, characterized in that: The honeycomb aperture of the protective cover (4) gradually decreases from the third acoustic resistance zone (403) to the first acoustic resistance zone (401), and the honeycomb aperture of the third acoustic resistance zone (403) is a through-hole structure, while the honeycomb aperture of the first acoustic resistance zone (401) is a blind hole or micro-hole structure.

5. A method for reducing echo problems in video conferencing, characterized in that: Includes the following steps: S1: Obtain the porosity of each of the first to third acoustic resistance zones in the directional microphone shield. With sound insulation layer density ,in Based on the porosity and density parameters, a frequency-related parameter is constructed. and the angle of incidence of sound waves Related gradient acoustic drag transfer function ; S2: Real-time acquisition of far-end speech reference signal output from local speaker. and mixed signals picked up by near-end directional microphones ; The gradient acoustic impedance transfer function in the frequency domain With mixed signals Perform convolution operations to compensate for the fixed acoustic distortion caused by the physical structure of the protective shield, and obtain the compensated signal. ; The far-end speech reference signal is calculated based on the cross-correlation function. With the compensated signal Dynamic delay parameters between ; S3: Based on the dynamic delay parameters For remote voice reference signal Perform time delay alignment to generate a time delay aligned reference signal. Construct a coupled adaptive filter and apply the gradient acoustic impedance transfer function. The echo estimation signal is generated by embedding the prior acoustic constraints into the update rules of the filter coefficients. ; S4: The compensated signal Subtract the echo estimation signal The error signal is obtained. The residual echo suppression threshold is dynamically configured based on the porosity ratio of each acoustic impedance zone, and the error signal is then processed based on this threshold. Nonlinear residual echo suppression processing is performed to ultimately output a clean near-end speech signal.

6. The method for reducing echo faults in video conferencing according to claim 5, characterized in that: In S1, the porosity and corresponding sound insulation layer density parameters of the three acoustic resistance zones of the microphone are first measured and collected. The two types of parameters are normalized to eliminate the differences in the dimensions and numerical magnitudes of the parameters, and the standardized normalized porosity and normalized sound insulation layer density are obtained. Based on the theory of sound wave propagation, and combined with the characteristics of porosity gradient distribution and density inverse correlation distribution, parameters such as acoustic resistance zone weighting coefficient, sound speed in air, axial length of acoustic resistance zone, imaginary unit, and cosine value of sound wave incident angle are incorporated to construct a frequency domain-angle domain joint gradient acoustic resistance transfer function that takes into account both sound wave frequency and incident angle. This allows for the quantification of the screening and attenuation effect of the protective shield structure on sound waves.

7. The method for reducing echo failure in video conferencing according to claim 5, characterized in that: In S2, a fast Fourier transform is first performed on the near-end mixed signal to convert the time-domain signal into a frequency-domain signal. Then, a frequency-domain convolution operation is performed with the gradient acoustic impedance transfer function to cancel the inherent distortion of the structure. Subsequently, an inverse fast Fourier transform is performed on the frequency-domain compensated signal to restore the time-domain compensated signal. Next, a cross-correlation operation is performed on the far-end reference signal and the time-domain compensated signal. By calculating the time delay value corresponding to the maximum value of the cross-correlation function, the index corresponding to the cross-correlation peak value is extracted to determine the dynamic time delay parameter in units of sampling points.

8. A method for reducing echo faults in video conferencing according to claim 5, characterized in that: In S3, the far-end reference signal is first shifted and aligned based on the obtained dynamic time delay parameters to obtain an aligned reference signal that matches the time sequence. Then, based on the normalized least mean square algorithm, the gradient acoustic impedance transfer function is used as a priori acoustic constraint and embedded in the update rule of the filter weights. Combined with the step size factor, regularization constant, and the L2 squared correlation parameter of the aligned reference signal, the filter weights are iteratively updated. Finally, the echo estimation signal is output by performing a convolution operation between the transpose of the filter weight vector and the aligned reference signal.

9. A method for reducing echo faults in video conferencing according to claim 5, characterized in that: In S4, the echo component subtraction operation is first performed, and the difference between the time-domain compensated signal and the echo estimation signal is calculated to obtain the error signal for preliminary removal of the main echo. Subsequently, based on the porosity ratio of the three acoustic impedance zones of the microphone, the residual echo suppression threshold is dynamically configured in conjunction with the threshold adjustment coefficient. Then, nonlinear amplitude determination processing is performed on the error signal. If the amplitude of the error signal is less than or equal to the dynamic threshold, it is set to zero. If the amplitude is greater than the threshold, the original signal value is retained. Finally, the processed near-end speech signal is output.

10. A method for reducing echo faults in video conferencing according to claim 5, characterized in that: The gradient acoustic resistance transfer function : in: The acoustic resistance zone weighting coefficient is positively correlated with the proportion of the axial length of each acoustic resistance zone, satisfying the following condition: ; Speed ​​of sound in air; : No. Axial length of each acoustic resistance zone; Imaginary unit, ; : The cosine of the incident angle of the sound wave, which characterizes the attenuation gain of the obliquely incident sound wave.

Images