A method for improving cardioid directivity of a microphone

Abstract

This invention relates to the field of audio processing and discloses a method for improving the cardioid directivity of a microphone, comprising the following steps: S1, performing frequency response equalization processing on a first audio signal acquired by an omnidirectional microphone and a second audio signal acquired by a dumbbell-shaped directional microphone to match the sensitivity and frequency response of the two microphones, wherein the equalization processing is implemented using an IIR or FIR filter; S2, adding the equalized first audio signal and the second audio signal with a 1:1 coefficient to generate a main beam audio signal pointing in the direction to be picked up. In this invention, by performing frequency response equalization processing on the signals acquired by the omnidirectional and dumbbell-shaped directional microphones, the reference beam is subsequently processed by a linear and nonlinear switchable adaptive filter, and the noise-reduced error signal is fed back to the filter to achieve dynamic iterative optimization of parameters. Finally, the time-domain high directivity audio is restored by inverse Fourier transform and bitstream splicing, which not only ensures the accuracy of beam generation but also makes noise cancellation more efficient.

Description

Technical Field

[0001] This invention relates to the field of audio processing, and more particularly to a method for improving the cardioid polar pattern of a microphone. Background Technology

[0002] In the field of audio pickup technology, microphone directivity is one of the core indicators determining pickup performance. Cardioid microphones, due to their ability to accurately pick up sound from the front and suppress background noise from the sides and rear, are widely used in various scenarios such as miniature smart terminals, portable audio devices, voice interaction devices, and handheld recording devices. As consumer electronic devices develop towards miniaturization, high integration, and high performance, the market has placed stringent demands on microphone modules: they must achieve high-level cardioid pickup with excellent backsound noise suppression and wide frequency response, while also meeting the integration requirements of small physical footprint and low hardware complexity, and adapting to the different sound quality and noise reduction switching needs of various scenarios such as high-definition voice calls and recording in noisy outdoor environments.

[0003] Currently, the traditional solutions for achieving microphone pickup directionality in the industry are mainly divided into two categories: unidirectional condenser microphone solutions and microphone linear array solutions. However, both types of solutions have obvious technical defects and are difficult to adapt to the comprehensive performance requirements of miniaturized integrated devices. 1. While the unidirectional condenser microphone solution boasts a wide frequency response of 100~15kHz and a simple hardware structure, its cardioid polar pattern is poor, with a -6dB beamwidth of 210°, resulting in ambiguous pickup directionality and easy pickup of irrelevant noise from the sides and rear. Furthermore, its rearward rejection ratio is no more than 15dB, indicating weak noise shielding capabilities, making it unsuitable for pickup in noisy environments. Additionally, its hardware footprint is 20... With a diameter of 20mm², the integration is relatively low, making it difficult to adapt to the hardware layout of micro-devices. 2. The microphone linear array solution achieves directional adjustment through the arrangement of multiple microphone arrays, achieving a backshort rejection ratio of 20-50dB and relatively superior directional performance. However, its performance is strongly correlated with the array size, requiring 3-8 microphones. More microphones result in better directional performance but lead to a significant narrowing of the frequency response (only 300-8kHz), sacrificing sound reproduction quality. Furthermore, this solution occupies a physical area of ​​1080-20150mm², resulting in bulky hardware and high circuit complexity. This not only increases the design and manufacturing costs of the device but also fails to meet the small-area integration requirements of miniature smart terminals and portable audio devices. Therefore, there is an urgent need for a new method to improve the cardioid polar pattern of microphones, which can achieve a combination of high-precision cardioid pickup, wide-frequency response sound quality reproduction and high-level back noise suppression in a small-area integrated hardware architecture, while supporting flexible switching between sound quality and noise reduction modes to adapt to the pickup needs of different application scenarios. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for improving the cardioid polar pattern of a microphone, thus solving the above problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for improving the cardioid polar pattern of a microphone, comprising the following steps: S1. Perform frequency response equalization processing on the first audio signal collected by the omnidirectional microphone and the second audio signal collected by the dumbbell-shaped directional microphone respectively, so that the two microphones are matched in terms of sensitivity and frequency response. The equalization processing is implemented using IIR or FIR filters. S2. Add the equalized first audio signal and the second audio signal with a 1:1 coefficient to generate the main beam audio signal pointing in the direction to be picked up; S3. Subtract the equalized first audio signal from the second audio signal by a 1:1 coefficient to generate a reference beam audio signal pointing in the opposite direction to the direction to be picked up. The main beam contains useful audio and noise in the range of -15dB, and the reference beam contains background noise and useful audio in the range of -15dB. S4. The main beam audio signal and the reference beam audio signal are processed by bitstream segmentation, with a segmentation overlap of 50% and a segmentation size of 2048 points. After adding a Hanning window, a fast Fourier transform is performed to convert to the frequency domain. In the frequency domain, the reference beam audio signal is subjected to adaptive filtering to obtain the filtered reference audio signal. S5. Subtract the filtered reference audio signal from the main beam audio signal in the frequency domain to perform noise cancellation processing and obtain the noise-reduced audio signal. S6. Feed the noise-reduced audio signal as an error signal to the adaptive filter, and adjust the parameters of the adaptive filter to minimize the error.

[0006] Preferably, the specific method for frequency response equalization processing includes: S11. Use a sweep frequency or wideband noise source to record omnidirectional microphones and dumbbell-shaped directional microphones in the direction of maximum sensitivity. S12. Calculate the sensitivity and frequency response differences of the two microphones respectively; S13. Based on the differences in sensitivity and frequency response, compensate one or both microphones to make them nearly identical in sensitivity and frequency response.

[0007] Preferably, the adaptive filtering process employs linear adaptive filtering or nonlinear adaptive filtering, wherein: the calculated frequency response width of linear adaptive filtering is 16kHz and the backward beam suppression is 25dB; the calculated frequency response width of nonlinear adaptive filtering is 12kHz and the backward beam suppression is 40dB.

[0008] Preferably, the adaptive filter used in the adaptive filtering process is one of the following: Kalman filter, least mean square filter, normalized least mean square filter, block LMS filter, fast block LMS filter, or segmented fast block LMS filter.

[0009] Preferably, the pointing direction of the reference beam is adjusted by the weight ratio of the omnidirectional microphone and the dumbbell-shaped directional microphone so that the reference beam points in the direction of the greatest noise, thereby achieving the best noise suppression effect.

[0010] Preferably, the omnidirectional microphone and the dumbbell-shaped directional microphone are integrated microphone modules, which have one omnidirectional microphone and one dumbbell-shaped directional microphone built in, and output two audio signals through a PDM interface.

[0011] Preferably, when the method is applied to a high dynamic range recording scenario, it further includes: A. The same microphone signal is split into two audio inputs with an amplitude difference of about 20dB, and then converted into a high-precision audio bitstream and a high-dynamic audio bitstream by two ADCs respectively. B. Simultaneously sample the two audio streams, calculate the amplitude difference in real time, and use an iterative approximation method to determine the convergence of the amplitude difference. C. Based on the convergence state of the amplitude difference and the signal amplitude, select a high-precision bitstream, a high-dynamic bitstream, or a mixture of both to output and synthesize a high-dynamic, high-signal-noise-ratio audio signal as the first audio signal or the second audio signal. D. Use the synthesized high dynamic range, high signal-to-noise ratio audio as the input audio signal for the method described in claim 1.

[0012] Preferably, it also includes S7, performing an inverse fast Fourier transform on the denoised audio signal, converting it back to the time domain after bitstream splicing and windowing, and outputting a highly directional audio signal.

[0013] Preferably, in step S4, the frame shift for bitstream segmentation is 1024 points, and the number of Fast Fourier Transform points is consistent with the segment size, which is 2048 points.

[0014] Preferably, in step S7, the -6dB beamwidth of the output highly directional audio signal is 90°~120°, and the backsuppression ratio is 25dB~40dB.

[0015] Beneficial effects This invention provides a method for improving the cardioid polar pattern of a microphone. Compared with existing technologies, it has the following advantages: In this invention, frequency response equalization is performed on the signals acquired by omnidirectional and dumbbell-shaped directional microphones. Through frequency sweeping noise reduction recording, difference calculation, and compensation calibration, the sensitivity and frequency response of the two microphones are highly matched. Then, the main beam and reference beam are combined using a 1:1 coefficient, ensuring the main beam accurately points to the direction of sound pickup and the reference beam points to the opposite noise region. Combined with 50% overlap bitstream segmentation, Hanning windowing, and Fourier transform frequency domain processing, the noise separation of the audio signal is more accurate. Subsequently, the reference beam is processed using a switchable linear and nonlinear adaptive filter, and the noise-reduced error... The difference signal is fed back to the filter to achieve dynamic iterative optimization of parameters. Finally, the time-domain highly directional audio is restored by inverse Fourier transform and bitstream splicing. This not only ensures the accuracy of beam generation, but also makes noise cancellation more efficient. The output signal has a -6dB beamwidth covering 90°~120° and a backward rejection ratio of 25dB~40dB. Compared with the traditional directional condenser microphone's 210° beamwidth and backward rejection ratio of no more than 15dB, the pickup directivity and noise shielding capability are greatly improved. This solves the technical problem that traditional solutions cannot simultaneously achieve small size, high directivity, strong noise reduction and wide frequency response. Attached Figure Description

[0016] Figure 1 This is a flowchart of a method for improving the cardioid polar pattern of a microphone, as proposed in this invention. Figure 2 Schematic diagram of linear and nonlinear filter beams; Figure 3 This is a schematic diagram of the frequency response of linear and nonlinear filters. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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.

[0018] Please see Figures 1-3 The present invention provides two technical solutions, specifically including the following embodiments: Example 1: A method for improving the cardioid polar pattern of a microphone includes the following steps: S1. Perform frequency response equalization processing on the first audio signal acquired by the omnidirectional microphone and the second audio signal acquired by the dumbbell-shaped directional microphone to match the sensitivity and frequency response of the two microphones. The equalization processing is implemented using IIR or FIR filters. The specific methods for frequency response equalization processing include: S11. Use a sweep frequency or wideband noise source to record omnidirectional microphones and dumbbell-shaped directional microphones in the direction of maximum sensitivity. S12. Calculate the sensitivity and frequency response differences of the two microphones respectively; S13. Based on the differences in sensitivity and frequency response, compensate one or two microphones to make them nearly identical in sensitivity and frequency response. S2. Add the equalized first audio signal and the second audio signal with a 1:1 coefficient to generate the main beam audio signal pointing in the direction to be picked up; S3. Subtract the equalized first audio signal from the second audio signal by a 1:1 coefficient to generate a reference beam audio signal pointing in the opposite direction to the direction to be picked up. The main beam contains useful audio and noise in the range of -15dB, and the reference beam contains background noise and useful audio in the range of -15dB. S4. The main beam audio signal and the reference beam audio signal are processed into separate bitstream segments with a 50% overlap and a segment size of 2048 points. After applying a Hanning window, a Fast Fourier Transform is performed to convert to the frequency domain. In the frequency domain, the reference beam audio signal is adaptively filtered to obtain the filtered reference audio signal. The adaptive filtering process uses either linear or nonlinear adaptive filtering. Specifically: the calculated frequency response width of the linear adaptive filter is 16kHz, and the backward beam suppression is 25dB; the calculated frequency response width of the nonlinear adaptive filter is 12kHz, and the backward beam suppression is 40dB. The adaptive filter used in the adaptive filtering process is a Kalman filter. The reference beam is selected from one of the following: a minimum mean square filter, a normalized minimum mean square filter, a block LMS filter, a fast block LMS filter, or a segmented fast block LMS filter. The pointing direction of the reference beam is adjusted by the weight ratio of the omnidirectional microphone and the dumbbell-shaped directional microphone to make the reference beam point in the direction of the maximum noise, so as to obtain the optimal noise suppression effect. The omnidirectional microphone and the dumbbell-shaped directional microphone adopt an integrated microphone module. The integrated microphone module has one omnidirectional microphone and one dumbbell-shaped directional microphone built in, and outputs two audio signals through the PDM interface. The frame shift of the bitstream segmentation processing is 1024 points, and the number of fast Fourier transform points is consistent with the segment size, which is 2048 points. S5. Subtract the filtered reference audio signal from the main beam audio signal in the frequency domain to perform noise cancellation processing and obtain the noise-reduced audio signal. S6. Feed the noise-reduced audio signal as an error signal to the adaptive filter, and adjust the parameters of the adaptive filter to minimize the error. S7. Perform inverse fast Fourier transform on the noise-reduced audio signal, and convert it back to the time domain after bitstream splicing and windowing to output a highly directional audio signal. The output highly directional audio signal has a -6dB beamwidth of 90°~120° and a backward suppression ratio of 25dB~40dB.

[0019] When the method is applied to high dynamic range recording scenarios, it also includes: A. The same microphone signal is split into two audio inputs with an amplitude difference of about 20dB, and then converted into a high-precision audio bitstream and a high-dynamic audio bitstream by two ADCs respectively. B. Simultaneously sample the two audio streams, calculate the amplitude difference in real time, and use an iterative approximation method to determine the convergence of the amplitude difference. C. Based on the convergence state of the amplitude difference and the signal amplitude, select a high-precision bitstream, a high-dynamic bitstream, or a mixture of both to output and synthesize a high-dynamic, high-signal-noise-ratio audio signal as the first audio signal or the second audio signal. D. Use the synthesized high dynamic range, high signal-to-noise ratio audio as the input audio signal for the method of claim 1.

[0020] Example 2: Based on Example 1, this design is optimized for small-area integrated hardware scenarios, adapting to applications such as micro smart terminals and portable audio devices that have stringent requirements for microphone module footprint and hardware integration. Currently, in traditional microphone pickup solutions, while directional condenser microphones (taking the 2016 electret condenser microphone as an example) have acceptable frequency response, their backward rejection ratio is no greater than 15dB, their -6dB beamwidth reaches 210°, resulting in poor directivity and noise suppression. Furthermore, they occupy an area of ​​2020mm², indicating low integration. While microphone linear arrays can achieve a backward rejection ratio of 20~50dB, their performance is strongly correlated with array size, requiring 3~8 microphones. More microphones result in better directivity but a narrower frequency response (only 300~8kHz), and their physical footprint reaches 1080~20mm². The 150mm² area is insufficient for the hardware layout requirements of miniature portable devices. This embodiment addresses the shortcomings of the traditional solution by making targeted optimizations, specifically including the following steps: S1. An integrated PDM interface dual-microphone module is used to complete audio signal acquisition. This module has one built-in omnidirectional microphone and one dumbbell-shaped directional microphone, with a physical footprint of 10. The 20mm² module directly outputs two raw audio signals (the first audio signal collected by the omnidirectional microphone and the second audio signal collected by the dumbbell-shaped directional microphone) synchronously through the PDM interface, without the need for additional signal conversion circuits, reducing hardware size and signal loss. The two raw audio signals output from the PDM interface are subjected to frequency response equalization processing. The equalization processing is implemented using an FIR filter to match the sensitivity and frequency response of the two microphones. The specific frequency response equalization processing steps are the same as S11-S13 in Example 1. After equalization, the frequency response of the two microphones is unified to the range of 100~16kHz (linear filtering mode) or 100~12kHz (non-linear filtering mode). S2. The equalized first audio signal and the second audio signal are added together with a 1:1 coefficient to generate a main beam audio signal pointing in the direction to be picked up. The sensitivity of the main beam in the direction to be picked up is the optimal value after the superposition of the two signals, and the basic -6dB beamwidth of the main beam is 120°. S3. Subtract the equalized first audio signal from the second audio signal by a 1:1 coefficient to generate a reference beam audio signal pointing in the opposite direction to the direction to be picked up. The main beam contains useful audio and noise in the range of -15dB, and the reference beam contains background noise and useful audio in the range of -15dB. The weight ratio of the omnidirectional microphone and the dumbbell-shaped directional microphone can be finely adjusted (within ±10%) to make the reference beam accurately point to the direction of the maximum noise in the actual use scenario of the device (such as the back and side ambient noise areas of portable devices), thereby improving the actual effect of noise suppression. S4. The main beam audio signal and the reference beam audio signal are processed into code stream segments with a segment overlap of 50% and a segment size of 2048 points. After adding a Hanning window, a fast Fourier transform is performed to convert to the frequency domain. In the frequency domain, the reference beam audio signal is subjected to switchable adaptive filtering. The filtering mode can be manually or automatically switched according to the sound quality and noise reduction requirements of the device's usage scenario to obtain the filtered reference audio signal: When the device is used in high-quality sound-priority scenarios such as high-definition voice calls and music pickup, linear adaptive filtering is selected, which is implemented using a normalized minimum mean square filter, with a calculated frequency response width of 16kHz and a backward beam suppression of 25dB, ensuring wide-frequency sound quality restoration; When the device is used in high-noise-priority scenarios such as outdoor noisy environments and public place recording, nonlinear adaptive filtering is selected, which is implemented using a block LMS filter, with a calculated frequency response width of 12kHz and a backward beam suppression of 40dB, enhancing the noise shielding effect of the rear and sides; S5. Subtract the filtered reference audio signal from the main beam audio signal in the frequency domain to perform noise cancellation processing and obtain the noise-reduced audio signal.

[0021] S6. Feed the noise-reduced audio signal as an error signal to the adaptive filter. Adjust the parameters of the adaptive filter through an iterative algorithm to minimize the amplitude of the error signal and achieve real-time dynamic optimization of the filter parameters.

[0022] S7. Perform an inverse fast Fourier transform on the denoised audio signal, and after bitstream splicing and windowing, convert it back to the time domain to output a highly directional audio signal. The performance of this signal can be switched according to the filtering mode. In linear filtering mode, the -6dB beamwidth is 120°, the backward rejection ratio is 25dB, and the frequency response is 100~16kHz. In nonlinear filtering mode, the -6dB beamwidth is 90°, the backward rejection ratio is 40dB, and the frequency response is 100~12kHz. The overall -6dB beamwidth covers 90°~120°, and the backward rejection ratio covers 25dB~40dB, matching the hardware performance of the integrated module.

[0023] When this method is applied to high dynamic range recording scenarios in small-area integrated devices, the following steps are included before step S1 above: A. The original audio signal output from the integrated microphone module is split into two audio inputs with an amplitude difference of about 20dB. These inputs are then converted from analog to digital by two high-precision ADCs to obtain a high-precision audio bitstream and a high-dynamic audio bitstream. The ADCs are integrated with the microphone module at the board level without increasing the physical footprint. B. Perform hardware-level synchronous sampling on the two audio streams, calculate the amplitude difference between the two streams in real time, and use an iterative approximation method to determine the convergence of the amplitude difference. The convergence threshold is set to ±1dB. C. Based on the convergence state of the amplitude difference and the real-time amplitude of the signal, automatically select a high-precision bitstream, a high-dynamic bitstream, or a mixture of both to output and synthesize a high-dynamic, high-signal-noise-ratio audio signal as the first or second audio signal of this method. D. Input the synthesized high dynamic range, high signal-to-noise ratio audio to step S1 for subsequent frequency response equalization and beamforming processing.

[0024] The following are the actual data results: Conclusion: The method in this embodiment is adapted to small-area integrated microphone modules of 1020mm². Compared with traditional microphone line arrays (1080~20150mm²) and conventional directional condenser microphones (2020mm²), it significantly reduces the physical footprint and improves hardware integration while achieving a better backward rejection ratio and more flexible pickup performance adjustment. Unlike microphone line arrays, it does not require sacrificing frequency response width to improve directivity. It perfectly balances the hardware layout limitations of micro smart terminals and portable audio devices with the dual requirements of sound quality and noise reduction in actual use, solving the technical problem that traditional solutions cannot simultaneously meet the requirements of small-area integration, high directivity, wide frequency response, and strong noise suppression.

[0025] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method of improving the cardioid directivity of a microphone, characterized by: Includes the following steps: S1. Perform frequency response equalization processing on the first audio signal collected by the omnidirectional microphone and the second audio signal collected by the dumbbell-shaped directional microphone respectively, so that the two microphones are matched in terms of sensitivity and frequency response. The equalization processing is implemented using IIR or FIR filters. S2. Add the equalized first audio signal and the second audio signal with a 1:1 coefficient to generate the main beam audio signal pointing in the direction to be picked up; S3. Subtract the equalized first audio signal from the second audio signal by a 1:1 coefficient to generate a reference beam audio signal pointing in the opposite direction to the direction to be picked up. The main beam contains useful audio and noise in the range of -15dB, and the reference beam contains background noise and useful audio in the range of -15dB. S4. The main beam audio signal and the reference beam audio signal are processed by bitstream segmentation, with a segmentation overlap of 50% and a segmentation size of 2048 points. After adding a Hanning window, a fast Fourier transform is performed to convert to the frequency domain. In the frequency domain, the reference beam audio signal is subjected to adaptive filtering to obtain the filtered reference audio signal. S5. Subtract the filtered reference audio signal from the main beam audio signal in the frequency domain to perform noise cancellation processing and obtain the noise-reduced audio signal. S6. Feed the noise-reduced audio signal as an error signal to the adaptive filter, and adjust the parameters of the adaptive filter to minimize the error.

2. The method of claim 1, wherein: The specific methods for frequency response equalization processing include: S11. Use a sweep frequency or wideband noise source to record omnidirectional microphones and dumbbell-shaped directional microphones in the direction of maximum sensitivity. S12. Calculate the sensitivity and frequency response differences of the two microphones respectively; S13. Based on the differences in sensitivity and frequency response, compensate one or both microphones to make them nearly identical in sensitivity and frequency response.

3. The method of claim 1, wherein: The adaptive filtering process employs either linear adaptive filtering or nonlinear adaptive filtering. Specifically, the calculated frequency response width of linear adaptive filtering is 16kHz, and the backward beam suppression is 25dB; the calculated frequency response width of nonlinear adaptive filtering is 12kHz, and the backward beam suppression is 40dB.

4. The method of claim 1, wherein: The adaptive filtering process employs one of the following adaptive filters: Kalman filter, least mean square filter, normalized least mean square filter, block LMS filter, fast block LMS filter, or segmented fast block LMS filter.

5. The method of claim 1, wherein: The pointing direction of the reference beam is adjusted by the weight ratio of the omnidirectional microphone and the dumbbell-shaped directional microphone so that the reference beam points in the direction of the greatest noise, thereby achieving the best noise suppression effect.

6. The method of claim 5, wherein: The omnidirectional microphone and the dumbbell-shaped directional microphone are integrated into a microphone module. The integrated microphone module has one omnidirectional microphone and one dumbbell-shaped directional microphone built in, and outputs two audio signals through the PDM interface.

7. The method of claim 1, wherein: When the method is applied to high dynamic range recording scenarios, it further includes: A. The same microphone signal is split into two audio inputs with an amplitude difference of about 20dB, and then converted into a high-precision audio bitstream and a high-dynamic audio bitstream by two ADCs respectively. B. Simultaneously sample the two audio streams, calculate the amplitude difference in real time, and use an iterative approximation method to determine the convergence of the amplitude difference. C. Based on the convergence state of the amplitude difference and the signal amplitude, select a high-precision bitstream, a high-dynamic bitstream, or a mixture of both to output and synthesize a high-dynamic, high-signal-noise-ratio audio signal as the first audio signal or the second audio signal. D. Use the synthesized high dynamic range, high signal-to-noise ratio audio as the input audio signal for the method described in claim 1.

8. The method of claim 1, wherein: It also includes step S7, performing an inverse fast Fourier transform on the denoised audio signal, and converting it back to the time domain after bitstream splicing and windowing to output a highly directional audio signal.

9. The method of claim 1, wherein: In step S4, the frame shift for bitstream fragmentation is 1024 points, and the number of Fast Fourier Transform points is consistent with the fragment size, which is 2048 points.

10. The method of claim 8, wherein: In step S7, the -6dB beamwidth of the output high directional audio signal is 90°~120°, and the backsuppression ratio is 25dB~40dB.

Images