Method for reducing outdoor pickup wind noise and outdoor pickup equipment thereof

Abstract

According to the method for reducing the outdoor pickup wind noise and the outdoor pickup equipment, the balance between the complexity and the effect of the scheme can be achieved through specific structural design and audio signal processing under the condition that only two microphones exist, and the system is more suitable for miniaturized voice communication equipment, occupies fewer hardware devices, and is high in practicability. And the requirement of reducing the wind noise in real-time voice communication can be met by using lower computing resources. In addition, an acoustic structure and an audio signal processing method are considered, an integrated scheme is formed, and the problem of adaptability is avoided.

Description

Technical Field

[0001] The invention belongs to the technical field of microphones, and in particular relates to a method for reducing wind noise in outdoor sound pickup and an outdoor sound pickup device thereof. Background Art

[0002] Wind noise is generated by the interaction between objects moving in the flow field, or by the interaction between fluids caused by the turbulent motion of the fluid itself. In the use of outdoor communication equipment, wind noise has been a problem that has long plagued academia and industry, especially in harsh outdoor environments with strong winds. Wind noise will interfere with normal voice signals, causing communication signals to be interfered with, affecting the effective communication of information. Existing wind noise reduction solutions have extremely high requirements for software and hardware, often using 6+1 microphone arrays or 8+1 microphone arrays, with high hardware costs and high signal processing computational complexity. Overly complex acoustic structure designs often make the microphone pickup hole farther away from the outside world, which reduces wind noise while seriously affecting the pickup of voice signals. In addition, acoustic structure and audio signal processing are often performed separately and cannot be organically combined. Summary of the invention

[0003] In view of this, an object of the present invention is to provide a method for reducing outdoor sound pickup wind noise and its outdoor sound pickup device and heating and cooling method, so as to overcome the above problems or at least partially solve or alleviate the above problems.

[0004] The present invention provides a method for reducing outdoor sound pickup wind noise, comprising the following steps:

[0005] S001: Pick up noisy audio signals through two microphones;

[0006] S002: Perform analog-to-digital conversion and encoding processing on the two signals;

[0007] S003: Pre-emphasize the two encoded audio signals to enhance the high-frequency components in the voice signals and reduce the influence of lip radiation;

[0008] S004: Determine the frame length and frame shift, and divide the non-stationary speech signal into frame signals with short-term stability through frame processing;

[0009] S005: Perform fast Fourier transform on the two signals, transform the time domain signals into the frequency domain to obtain two frequency domain signals;

[0010] S006: Eliminate its influence on different frequencies of audio signals through transmission matrix;

[0011] S007: Process the two frequency domain signals through the filter coefficient matrix to obtain a speech channel and a noise channel;

[0012] S008: combining the speech channel and the noise channel as the desired signal and the error signal respectively with the adaptive filtering algorithm, estimating the optimal step length of the filter using the two-channel signals, controlling the weight coefficient of the adaptive filter using the estimated optimal step length, and updating the weight of each frequency component through continuous iteration;

[0013] S009: Perform inverse Fourier transform processing on the frequency domain signal processed by the algorithm, and finally obtain a noise-reduced speech signal.

[0014] The present invention also has the following optional features.

[0015] Optionally, in step S004, the frame length of the frame signal is 30 ms, the frame shift is 15 ms, and the width of the frame signal is 20 ms-30 ms.

[0016] Optionally, in step S006, a transmission matrix is ​​obtained from the porous metamaterial in the microphone. The transmission matrix is ​​determined by the material and structure of the material and can be obtained through testing or theoretical calculation.

[0017] Optionally, in step S007, the filter coefficient matrix may be changed according to actual needs. The model used here is a back-to-back cardioid beam, using a first-order forward-backward difference filter.

[0018] Optionally, in step S008, the adaptive filtering algorithm is an LMS (Least Mean Square Algorithm) algorithm, and the LMS algorithm iteratively updates the weight coefficient of the filter to minimize the mean square error between the desired signal and the filter output.

[0019] The present invention also proposes an outdoor sound pickup device, including a main body, on which two microphones are arranged, a porous metamaterial is arranged in front of the sound pickup hole of each microphone, a waterproof sound-permeable membrane and a buffer foam are arranged between the porous metamaterial and the sound pickup hole of the microphone, and a rubber sleeve is also arranged on the outside of the microphone.

[0020] The method for reducing wind noise in outdoor sound pickup and the outdoor sound pickup device of the present invention achieve a balance between the complexity and effect of the solution through specific structural design and audio signal processing when there are only two microphones. The system is more suitable for miniaturized voice communication equipment, occupies fewer hardware devices, and uses lower computing resources to meet the requirements of wind noise reduction in real-time voice communication. In addition, the present application takes into account both the acoustic structure and the audio signal processing method to form an integrated solution, avoiding adaptability issues. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is the signal waveform diagram before and after pre-emphasis processing;

[0022] Figure 2 This is the two-way signal diagram of the 160th frame after framing and windowing;

[0023] Figure 3 The 160th frame signal diagram for converting the signal into the frequency domain;

[0024] Figure 4 This is the 160th frame signal after eliminating the influence of porous metamaterial;

[0025] Figure 5 This is the two-way signal diagram of the 160th frame after passing through the filter coefficient matrix;

[0026] Figure 6 This is the 160th frame signal diagram before and after LMS adaptive filtering;

[0027] Figure 7 The final processed result and the time domain and frequency domain comparison of the signals picked up by the two microphones are shown below.

[0028] Figure 8 A schematic diagram of the microphone layout of the outdoor sound pickup device of the present invention;

[0029] Fig. 9 for Figure 8 Schematic diagram of the structure of the microphone.

[0030] In the above picture: 1. Main body; 2. Microphone; 3. Porous metamaterial; 4. Waterproof sound-transmitting membrane; 5. Buffering foam; 6. Rubber cover.

[0031] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. DETAILED DESCRIPTION

[0032] The embodiment of the present invention proposes a method for reducing wind noise in outdoor sound pickup, comprising the following steps: S001: picking up noisy audio signals through two microphones; S002: performing analog-to-digital conversion and encoding processing on the two signals; S003: performing pre-emphasis processing on the encoded two audio signals to enhance the high-frequency components in the voice signals and reduce the influence of lip radiation; S004: determining the frame length and frame shift, and dividing the non-stationary voice signal into frame signals with short-term stability through frame processing; S005: performing fast Fourier transform on the two signals, transforming the time domain signals into the frequency domain to obtain two frequency domain signals; S006: Eliminate its influence on different frequencies of the audio signal through the transmission matrix; S007: Process the two frequency domain signals through the filter coefficient matrix to obtain the voice channel and the noise channel; S008: Combine the voice channel and the noise channel as the expected signal and the error signal respectively with the adaptive filtering algorithm, use the two-channel signals to estimate the optimal step size of the filter, use the estimated optimal step size to control the weight coefficient of the adaptive filter, and update the weight of each frequency component through continuous iteration; S009: Perform inverse Fourier transform processing on the frequency domain signal processed by the algorithm to finally obtain the denoised voice signal.

[0033] In step S001, a porous metamaterial is placed in front of the sound pickup holes of two microphones, and the two microphones pick up the noisy audio signal y 1 and 2 .

[0034] In step S002, the audio signal y is converted by an analog-to-digital conversion chip and a coding chip. 1 and 2 Perform analog-to-digital conversion and encoding processing to obtain audio signals y1 and y2. The default microphone here is an analog microphone. If it is a digital microphone, there is no analog-to-digital conversion process. The models of the analog-to-digital conversion chip and encoding chip used in this processing are ES7210 and ES8311 respectively.

[0035] In step S003, the encoded two-way audio signal is preprocessed. First, pre-emphasis is performed to enhance the high-frequency components in the speech signal and reduce the influence of lip radiation. The signals picked up by the two microphones to be processed are y1 and y2 respectively, and the pre-emphasized signals are y1e and y2e.

[0036] Figure 1 The signal waveforms of the two microphones before and after processing.

[0037] In step S004, the frame length and frame shift are determined, and the non-stationary speech signal is divided into frame signals with a width of 20ms-30ms and short-term stability through framing processing; the frame length of the frame signal is set to 30ms and the frame shift is set to 15ms. The framed signal is windowed to concentrate the energy on the main lobe to avoid frequency domain energy leakage. An audio file is divided into multiple frames according to the audio duration, frame length and frame shift. In order to more clearly understand the idea of ​​this algorithm, the subsequent frequency domain signal processing uses the 160th frame to illustrate. There is nothing special about the selection of this frame. It is just a random selection of a frame for the convenience of explanation.

[0038] Figure 2 This is the time domain diagram of the two-way signal at the 160th frame after processing in this step.

[0039] In step S005, the two signals after step 4 are subjected to fast Fourier transform, and the time domain signals are transformed into the frequency domain to obtain two frequency domain signals Y 1e (k) and Y 2e (k) Y 1e (k) and Y 2e (k) Data type is matrix, Y 1e (k)[i] and Y 2e (k)[i] is the i-th frame signal corresponding to it,

[0040]

[0041]

[0042] Figure 3 This is the frequency domain diagram of the 160th frame of the two signals after being processed in this step.

[0043] In step S006, a transmission matrix K is obtained from the porous metamaterial to eliminate its influence on different frequencies of the audio signal.

[0044] The transmission matrix is ​​determined by the material and structure of the material. The sound signal intensity before passing through the porous material is x_in, and the sound signal after passing through the porous material is x_out = K*x_in. The transmission matrix is ​​a function of frequency and can be calculated by testing or theory. The calculation method of the transmission matrix in this case is as follows, which takes into account the thermal viscosity under small pore size:

[0045]

[0046] Z 0 =ρc, S = πa 2 ,

[0047]

[0048] Where ω is the angular frequency ω = 2πf, L is the thickness of the porous metamaterial, ρ is the air density, μ is the dynamic viscosity, Cp is the heat capacity at constant pressure, k is the thermal conductivity, c is the speed of sound, S is the cross-sectional area of ​​the hole, a is the radius of the hole, J2 and J0 are the 2nd order first kind Bessel function and the 0th order first kind Bessel function, respectively. γ is the specific heat rate, kv is the viscous wave number, kth is the thermal wave number, k0 is the wave number, Z0 is the air acoustic impedance, Yv is the average value of the scalar temperature field, Yth is the average value of the scalar viscosity field, Zc is the characteristic acoustic impedance including the thermoviscous effect, and kc is the modal wave number. In this case, c = 343m / s. Considering that the effective frequency f of the audio signal in this case is in the range of 0-4kHz, the range of ω is corresponding to it. The relationship between ω and f is ω = 2πf, L = 1mm, Z0 = 400Pa*s / m, a = 0.1mm, μ = 1.81*10^-5pa*s, Cp = 1005.4J / (kg*K), S = 0.0314mm^2, γ = 1.4. The remaining quantities can be obtained by the above formula. After finding the K matrix, invert it to get K -1 , K -1 Respectively with two signal Y 1e and Y 1e Perform matrix multiplication to finally get the result Y that eliminates the effect of porous materials on the signal 1 and Y 2 ,Right now

[0049]

[0050] Figure 4 This is the spectrum diagram of the 160th frame before and after the two-way signal processing.

[0051] In step S007, the filter coefficient matrix H * Process the two frequency domain signals Y1 and Y2 obtained in the above steps to obtain Y F (k) and Y B (k) Y F (k) contains audio signal and noise signal components, Y B (k) mainly contains the noise signal component, and the filter coefficient matrix H * It can be changed according to actual needs. The model used here is a back-to-back cardioid beam, using a first-order forward and backward difference filter, and the matrix H * for in, Through matrix operations Get the required voice channel signal Y F (k) and the noisy channel signal Y B(k), as above, ω is the angular frequency ω = 2πf, f ranges from 0 to 4kHz, τ 0 =0.16ms.

[0052] Figure 5 is the spectrum of the 160th frame of the two signals.

[0053] In step S008, the voice channel Y F (k) and the noise channel Y B (k) are combined with the adaptive filtering algorithm as the expected signal and error signal respectively, and the optimal step length of the filter is estimated using the two-channel signals. The weight coefficient W(k) of the adaptive filter is controlled by the estimated optimal step length. By continuously iteratively updating the weights of each frequency component, the spectrum of the interference signal will continue to offset as the number of iterations increases, thereby further suppressing Y F (k) is the effect of residual noise, and E(k) is a parameter that characterizes the error estimation. The adaptive filtering algorithm used is the LMS (Least Mean Square Algorithm) algorithm, which iteratively updates the weight coefficients of the filter to minimize the mean square error between the desired signal and the filter output.

[0054] Among them, the weight vector is W(k)=[W 0 (k),W 1 (k),W 2L-1 (k)] T , L represents the frame sampling length, which is 241, the number of sampling points corresponding to a 30ms frame length at an 8kHz sampling rate. The frequency domain weight coefficient vector update formula of the adaptive filter is: W(k+1)=W(k)+μ B *Y B (k)*e * (k), where μ B is a variable step factor, the starting value W is set to 0, the step factor is set to 0.01, and the iteration order is set to 20.

[0055] Figure 6 The final signal spectrum is shown in Figure 2. The dark color is the signal after noise reduction, and the light color is the signal before LMS adaptive filtering.

[0056] In step S009, the frequency domain signal Y processed by the algorithm is out (k) is processed later to finally obtain the de-noised speech signal s(t). This process is time-frequency conversion. From frequency domain signal to time domain signal, the inverse Fourier transform is Get the final processed result.

[0057] Figure 7The final processed result and the time domain and frequency domain comparison of the signals picked up by the two microphones show that the noise has been effectively suppressed. The spectrum of the speech signal can be seen more clearly, and the spectrum when people are not speaking is also relatively clean.

[0058] like Figure 8 and Fig. 9 As shown, the present invention also proposes an outdoor sound pickup device, including a main body 1, on which two microphones 2 are arranged, a porous metamaterial 3 is arranged in front of the sound pickup hole of each microphone 2, a waterproof sound-permeable membrane 4 and a buffer foam 5 are arranged between the porous metamaterial 3 and the sound pickup hole of the microphone 2, and a rubber sleeve 6 is also arranged on the outside of the microphone 2.

[0059] The main body 1 is a small outdoor sound pickup device such as a mobile phone. Two sound holes 101 are arranged near the edge of the main body 1. Microphones 2 are installed inside the two sound holes 101. A porous metamaterial 3 is arranged between the sound pickup hole of the microphone 2 and the sound hole 101. The porous metamaterial 3 can effectively reduce the air flow velocity in front of the sound pickup hole, modulate its flow mode, reduce the occurrence of turbulence, and prevent the air flow from exciting its own vibration. Isolation and buffering protection is performed.

[0060] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims. The components and structures not described in detail in this embodiment are well-known components and common structures or common means in the industry, and are not described here one by one.

Claims

1. A method for reducing outdoor wind noise, characterized in that: The following steps are involved: S001: Pick up noisy audio signals through two microphones; S002: Perform analog-to-digital conversion and encoding processing on the two signals; S003: Pre-emphasize the two encoded audio signals to enhance the high-frequency components in the voice signals and reduce the influence of lip radiation; S004: Determine the frame length and frame shift, and divide the non-stationary speech signal into frame signals with short-term stability through frame processing; S005: Perform fast Fourier transform on the two signals, transform the time domain signals into the frequency domain to obtain two frequency domain signals; S006: Eliminate its influence on different frequencies of audio signals through transmission matrix; S007: Process the two frequency domain signals through the filter coefficient matrix to obtain a speech channel and a noise channel; S008: combining the speech channel and the noise channel as the desired signal and the error signal respectively with the adaptive filtering algorithm, estimating the optimal step length of the filter using the two-channel signals, controlling the weight coefficient of the adaptive filter using the estimated optimal step length, and updating the weight of each frequency component through continuous iteration; S009: Perform inverse Fourier transform processing on the frequency domain signal processed by the algorithm, and finally obtain a noise-reduced speech signal.

2. The method for reducing outdoor wind noise pickup according to claim 1, characterized in that: In step S004, the frame length of the frame signal is 30 ms, the frame shift is 15 ms, and the width of the frame signal is 20 ms-30 ms.

3. The method for reducing outdoor wind noise pickup according to claim 1, characterized in that: In step S006, a transmission matrix is ​​obtained from the porous metamaterial in the microphone. The transmission matrix is ​​determined by the material and structure of the material and can be obtained through testing or theoretical calculation.

4. The method for reducing outdoor wind noise pickup according to claim 1, characterized in that: In step S007, the filter coefficient matrix can be changed according to actual needs. The model used here is a back-to-back cardioid beam, using a first-order forward-backward difference filter.

5. The method for reducing outdoor wind noise pickup according to claim 1, characterized in that: In step S008, the adaptive filtering algorithm is an LMS (Least Mean Square Algorithm) algorithm, which iteratively updates the weight coefficients of the filter to minimize the mean square error between the desired signal and the filter output.

6. An outdoor sound pickup device, characterized in that: The invention comprises a main body (1), on which two microphones (2) are arranged, a porous metamaterial (3) is arranged in front of the sound pickup hole of each microphone (2), a waterproof sound-permeable membrane (4) and a buffer foam (5) are arranged between the porous metamaterial (3) and the sound pickup hole of the microphone (2), and a rubber sleeve (6) is also arranged on the outside of the microphone (2).

Images