Microphone array wind noise reduction optimization method, apparatus and computer-readable storage medium

Abstract

This invention relates to a method, apparatus, and computer-readable storage medium for optimizing wind noise reduction in a microphone array. The wind noise reduction optimization method includes setting a reference microphone on the microphone array and setting a physical wind noise reduction device on the reference microphone; the reference microphone and each main microphone simultaneously acquiring ambient audio signals; calculating the mean square error of the signal acquired by the main microphone relative to the signal acquired by the reference microphone; calculating the weighting coefficient of the main microphone based on the mean square error; and synchronously optimizing the signals acquired by each main microphone based on the weighting coefficient and the signal acquired by the reference microphone. In this invention, by setting a physical wind noise reduction device on the reference microphone, the reference microphone can provide a more stable reference signal for the noise reduction algorithm in windy conditions; in addition, calculating the weighting coefficient of the main microphone based on the mean square error can further improve the noise reduction effect of the microphone array, effectively reducing or eliminating wind noise.

Description

Technical Field

[0001] This invention belongs to the field of microphone array wind noise reduction technology, and relates to a microphone array wind noise reduction optimization method, device and computer-readable storage medium. Background Technology

[0002] Microphone arrays are widely used in speech recognition, voice commands, and voice communication. However, when installed in environments with fans or air conditioning vents, they face wind noise interference, which degrades the quality of the voice signal and affects the performance of the voice processing system and the user experience. Traditionally, this problem is solved by covering all microphones with a layer of windproof fleece, but this significantly increases equipment costs. Therefore, it is necessary to find a more economical and effective solution. Summary of the Invention

[0003] In view of the shortcomings of the prior art, the technical problem to be solved by the present invention is to provide a microphone array wind noise reduction optimization method, device and computer-readable storage medium.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for optimizing wind noise reduction in microphone arrays includes the following steps:

[0006] S100. Set a reference microphone on the microphone array and set a physical wind noise reduction device on the reference microphone;

[0007] S200: Define microphones other than the reference microphone in the microphone array as main microphones, and simultaneously collect ambient audio signals from the reference microphone and each main microphone.

[0008] S300, Calculate the mean square error of the signal acquired by each main microphone relative to the signal acquired by the reference microphone;

[0009] S400: Calculate the weighting coefficient of each main microphone based on the mean square error;

[0010] S500 performs wind noise reduction optimization on the signals collected by each main microphone simultaneously, based on the weighting coefficient of the main microphone and the signal collected by the reference microphone.

[0011] Furthermore, in step S100, the method for setting the reference microphone is as follows:

[0012] Specify a microphone as the reference microphone in the microphone array, or set a separate microphone as the reference microphone.

[0013] Furthermore, in step S100, the physical wind noise reduction device includes windproof down and / or windproof cover.

[0014] Furthermore, after performing step S200, the following steps are also performed:

[0015] S210. Store the audio signal data collected by the reference microphone and each main microphone as a time-domain signal sequence;

[0016] S220. The time-domain signal sequence is subjected to denoising, filtering and amplification to obtain preprocessed audio data;

[0017] S230. Extract time-domain or time-frequency domain features from the preprocessed audio data, and use the extracted features to characterize the signal acquired by the microphone.

[0018] Furthermore, in step S400, calculating the weighting coefficients of each main microphone based on the mean square error includes the following sub-steps:

[0019] S410. Normalize the mean square error of each main microphone to obtain the normalized weight of each main microphone.

[0020] S420. Calculate the initial weights of each main microphone based on their normalized weights.

[0021] S430. Adjust the initial weights of each main microphone proportionally to obtain the weight coefficients of each main microphone.

[0022] Furthermore, in step S410, the formula for normalization is as follows:

[0023]

[0024] Where i represents the sequence number of the main microphone; d_norm[i] represents the normalized weight of the i-th main microphone; d_mse[i] represents the mean square error of the i-th main microphone; d_min represents the minimum mean square error among all main microphones; and d_max represents the maximum mean square error among all main microphones.

[0025] In step S420, the formula for calculating the initial weights is as follows:

[0026]

[0027] Where w[i] represents the initial weight of the i-th main microphone; N represents the total number of main microphones;

[0028] In step S430, the formula for proportionally adjusting the initial weights is as follows:

[0029]

[0030] Where w[i]' represents the weight coefficient of the i-th main microphone.

[0031] Furthermore, in step S500, the formula for optimizing the wind noise reduction of the signal acquired by the main microphone is as follows:

[0032] y[i]=(1-w[i]')×x[i]+w[i]'×x_ref_processed

[0033] Where y[i] represents the signal obtained after wind noise reduction optimization of the i-th main microphone; x[i] represents the signal collected by the i-th main microphone; and x_ref_processed represents the signal collected by the reference microphone.

[0034] A microphone array wind noise reduction optimization device, comprising

[0035] A microphone array for acquiring ambient audio signals, the microphone array including a reference microphone and multiple main microphones, the reference microphone being equipped with a physical wind noise reduction device;

[0036] The mean square error calculation module is used to calculate the mean square error of the signals acquired by each main microphone relative to the signals acquired by the reference microphone.

[0037] The weight calculation module is used to calculate the weight coefficients of each main microphone based on the mean square error; and

[0038] The wind noise reduction optimization module is used to optimize the wind noise reduction of the signal collected by the main microphone based on the weighting coefficient of the main microphone and the signal collected by the reference microphone.

[0039] Furthermore, it also includes a preprocessing module, which is used to denoise, filter and amplify the audio signals collected by the reference microphone and each main microphone, and then extract their time domain features or time-frequency domain features, and use the extracted features to characterize the signals collected by the microphones.

[0040] A computer-readable storage medium having a computer program stored thereon, which, when loaded and executed by a processor, implements a method for optimizing wind noise reduction in a microphone array.

[0041] In this invention, by setting a physical wind noise reduction device on the reference microphone, the reference microphone can provide a more stable reference signal for the noise reduction algorithm in windy conditions, thus positively impacting the noise reduction effect of the microphone array. Furthermore, improvements to the weighting coefficient calculation method and the noise reduction algorithm can further enhance the noise reduction effect of the microphone array, effectively reducing or eliminating wind noise caused by wind in the microphone array and improving the quality of the acquired audio signal. Attached Figure Description

[0042] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0043] Figure 1 This is a flowchart of one embodiment of the microphone array wind noise reduction optimization method of the present invention.

[0044] Figure 2 This is a flowchart illustrating the process of processing the audio signals captured by the microphone.

[0045] Figure 3 This is a flowchart for calculating the weighting coefficients based on the mean square error.

[0046] Figure 4 This is a structural block diagram of one embodiment of the microphone array wind noise reduction optimization device of the present invention.

[0047] Figure 5 This is a block diagram of the weight calculation module.

[0048] The meanings of the labels in the attached diagram are as follows:

[0049] Microphone array - 100; Mean square error calculation module - 200; Weight calculation module - 300; Normalization processing unit - 310; Initial weight calculation unit - 320; Weight adjustment unit - 330; Wind noise reduction optimization module - 400; Preprocessing module - 500. Detailed Implementation

[0050] The following specific examples illustrate the implementation of the present invention. The illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0051] Please see Figure 1 , Figure 1 This is a flowchart illustrating one embodiment of a microphone array wind noise reduction optimization method according to the present invention. The microphone array wind noise reduction optimization method of this embodiment includes the following steps:

[0052] S100. Set a reference microphone on the microphone array and install a physical wind noise reduction device on the reference microphone. When setting the reference microphone, any microphone in the microphone array can be designated as the reference microphone, or an additional microphone can be added to the microphone array as the reference microphone. The physical wind noise reduction device can be a windproof fleece wrapped around the reference microphone and a windproof cover surrounding the fleece. Windproof fleece is a common noise reduction method used in microphone arrays; its biggest advantage is that it can effectively reduce or eliminate wind noise caused by wind blowing, thereby improving the quality of the voice signal. Of course, other devices that can reduce the impact of wind noise on the microphone can also be used as the physical wind noise reduction device.

[0053] S200: Microphones other than the reference microphone in the microphone array are defined as main microphones. The reference microphone and each main microphone simultaneously acquire ambient audio signals. Since the signals acquired by the main microphones and the reference microphones need to be compared to calculate the mean square error and to perform normalization, the signals acquired by the main microphones and the reference microphones must be aligned on the time axis. After the reference microphones and each main microphone acquire the ambient audio signals, the audio signals can be processed to initially remove noise and improve signal quality.

[0054] Please see Figure 2 The steps for processing the audio signal captured by the microphone are as follows:

[0055] S210. Store the audio signal data collected by the reference microphone and each main microphone as a time-domain signal sequence.

[0056] S220. The time-domain signal sequence is preprocessed to obtain preprocessed audio data. The preprocessing process includes denoising, filtering, and amplification. Common digital signal processing techniques can be used for preprocessing, such as wavelet denoising, notch filtering, and adaptive filtering, to improve the signal-to-noise ratio and reduce the impact of environmental noise.

[0057] S230. Extract features from the preprocessed audio data to characterize the signal acquired by the microphone for use in subsequent wind noise reduction optimization algorithms. The extracted features can be time-domain features, time-frequency domain features, etc.

[0058] S300. Based on the extracted features, calculate the mean square error (MSE) of the signal acquired by each main microphone relative to the signal acquired by the reference microphone. Assume the microphone array contains N main microphones, and the signal acquired by the i-th main microphone is x[i]; the signal acquired by the reference microphone after processing with a windproof velour and windscreen is x_ref_processed. Then, the formula for calculating the MSE of the signal acquired by the i-th main microphone relative to the signal acquired by the reference microphone is:

[0059] d_mse[i]=MSE(x[i],x_ref_processed)

[0060] Where i represents the sequence number of the main microphone; d_mse[i] represents the mean square error of the i-th main microphone; and MSE() is the mean square error function.

[0061] S400: Calculate the weighting coefficients of each main microphone based on the mean square error.

[0062] Please see Figure 3 This step may specifically include the following sub-steps:

[0063] S410. Normalize the mean square error of each main microphone to obtain the normalized weight of each main microphone. The formula for normalization is as follows:

[0064]

[0065] Where d_norm[i] represents the normalized weight of the i-th master microphone; d_min represents the minimum mean square error among all master microphones; and d_max represents the maximum mean square error among all master microphones. Normalization ensures that the normalized weights of all master microphones are within the range of [0,1]. The smaller the normalized weight of a master microphone, the greater its proportion in the final output of the microphone array.

[0066] S420. Calculate the initial weights for each main microphone based on its normalized weights. The formula for calculating the initial weights is as follows:

[0067]

[0068] Where w[i] represents the initial weight of the i-th main microphone; N represents the total number of main microphones.

[0069] S430. The initial weights of each main microphone are adjusted proportionally to obtain the weight coefficients for each main microphone. After adjustment, the sum of the weight coefficients of all main microphones is "1" to simplify the subsequent wind noise reduction optimization algorithm. The formula for proportionally adjusting the initial weights of the main microphones is as follows:

[0070]

[0071] Where w[i]' represents the weight coefficient of the i-th main microphone.

[0072] S500 performs wind noise reduction optimization synchronously on the signals acquired by each main microphone based on the weighting coefficient of the main microphone and the signal acquired by the reference microphone. During wind noise reduction optimization, it is necessary to ensure that the reference microphone and each main microphone are processed within the same time window to maintain synchronization. The formula for wind noise reduction optimization of the signals acquired by the main microphones is:

[0073] y[i]=(1-w[i]')×x[i]+w[i]'×x_ref_processed

[0074] Where y[i] represents the signal obtained after wind noise reduction optimization of the i-th master microphone. After calculating the output of each master microphone after wind noise reduction optimization, the summation of the outputs of all master microphones gives the output of the microphone array after wind noise reduction optimization.

[0075] In this embodiment, after selecting a reference microphone in the microphone array, windproof devices such as windproof felt are added to the reference microphone. This allows the reference microphone to maintain a relatively stable output signal in windy conditions, thereby providing a more stable reference signal for the noise reduction algorithm and positively impacting the noise reduction effect of the microphone array. Furthermore, improvements to the weighting coefficient calculation method and the noise reduction algorithm can further enhance the noise reduction effect of the microphone array, effectively reducing or eliminating wind noise caused by wind in the microphone array and improving the quality of the acquired audio signal.

[0076] Please see Figure 4 This is a structural block diagram of an embodiment of a microphone array wind noise reduction optimization device according to the present invention. This embodiment of the microphone array wind noise reduction optimization device includes a microphone array 100, a mean square error calculation module 200, a weight calculation module 300, and a wind noise reduction optimization module 400. Of course, in order to improve the signal-to-noise ratio of the signals acquired by the microphones and better remove noise, the microphone array wind noise reduction optimization device may also include a preprocessing module 500.

[0077] The microphone array 100 is used to collect ambient audio signals. The microphone array 100 consists of multiple microphones arranged in a specific geometric layout. One microphone in the microphone array 100 is designated as a reference microphone, and a physical wind noise reduction device is installed on the reference microphone; the remaining microphones in the microphone array 100 after removing the reference microphone are all defined as main microphones.

[0078] The preprocessing module 500 is used to denoise, filter, and amplify the audio signals acquired by the reference microphone and each main microphone, and then extract their time-domain or time-frequency domain features. The extracted features are used to characterize the signals acquired by the microphones. The preprocessing module 200 stores the audio signal data acquired by the reference microphone and each main microphone as a time-domain signal sequence, and performs denoising, filtering, and amplification on the time-domain signal sequence to obtain preprocessed audio data to improve the signal-to-noise ratio and reduce the influence of environmental noise; and extracts time-domain features, time-frequency domain features, and other features from the preprocessed audio data to characterize the signals acquired by the microphones for subsequent wind noise reduction optimization algorithms.

[0079] The mean square error calculation module 200 is used to calculate the mean square error of the signals acquired by each main microphone relative to the signals acquired by the reference microphone. The formula for calculating the mean square error d_mse[i] is:

[0080] d_mse[i]=MSE(x[i],x_ref_processed)

[0081] The weight calculation module 300 is used to calculate the weight coefficient of each main microphone based on the mean square error. First, the mean square error of each main microphone can be normalized to obtain a normalized weight for each main microphone; then, the initial weight of each main microphone can be calculated based on its normalized weight; finally, the initial weight of each main microphone can be adjusted. The weight calculation module 300 may include a normalization processing unit 310, an initial weight calculation unit 320, and a weight adjustment unit 330.

[0082] The normalization processing unit 310 is used to normalize the mean square error of each main microphone to obtain the normalized weight of each main microphone. The formula for normalization is as follows:

[0083]

[0084] The initial weight calculation unit 320 is used to calculate the initial weight of each main microphone based on its normalized weight. The formula for calculating the initial weight is as follows:

[0085]

[0086] The weight adjustment unit 330 is used to proportionally adjust the initial weights of each main microphone to obtain the weight coefficient of each main microphone. After adjustment, the sum of the weight coefficients of all main microphones is "1", which simplifies the subsequent wind noise reduction optimization algorithm. The formula for proportionally adjusting the initial weights of the main microphones is as follows:

[0087]

[0088] The wind noise reduction optimization module 400 is used to perform wind noise reduction optimization on the signal acquired by the main microphone based on the weighting coefficient of the main microphone and the signal acquired by the reference microphone. The formula for wind noise reduction optimization of the signal acquired by the main microphone is:

[0089] y[i]=(1-w[i]')×x[i]+w[i]'×x_ref_processed

[0090] In this embodiment, windproof devices such as windproof felt are added to the reference microphone, enabling it to provide a more stable reference signal for the noise reduction algorithm in windy conditions, thereby improving the noise reduction effect of the microphone array 100. Furthermore, improvements to the weighting coefficient calculation method and the noise reduction algorithm can further enhance the noise reduction effect of the microphone array 100, effectively reducing or eliminating wind noise caused by wind in the microphone array 100 and improving the quality of the acquired audio signal.

[0091] This invention also discloses a computer-readable storage medium. A preferred embodiment of this computer-readable storage medium stores a computer program that, when loaded and executed by a processor, fully or partially implements the microphone array wind noise reduction optimization method of the above embodiments. In this embodiment, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof. The details of the method implemented when the computer program is loaded and executed by the processor are described in the foregoing embodiments and will not be repeated here.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for optimizing wind noise reduction in microphone arrays, characterized in that, Includes the following steps: S100. Set a reference microphone on the microphone array and set a physical wind noise reduction device on the reference microphone; S200: Define microphones other than the reference microphone in the microphone array as main microphones, and simultaneously collect ambient audio signals from the reference microphone and each main microphone. S300, Calculate the mean square error of the signal acquired by each main microphone relative to the signal acquired by the reference microphone; S400: Calculate the weighting coefficient of each main microphone based on the mean square error; S500 performs wind noise reduction optimization on the signals collected by each main microphone simultaneously, based on the weighting coefficient of the main microphone and the signal collected by the reference microphone. In step S400, calculating the weighting coefficients of each main microphone based on the mean square error includes the following sub-steps: S410. Normalize the mean square error of each main microphone to obtain the normalized weight of each main microphone. S420. Calculate the initial weights of each main microphone based on their normalized weights. The formula for calculating the initial weights is as follows: Where i represents the sequence number of the main microphone; w[i] represents the initial weight of the i-th main microphone; d_norm[i] represents the normalized weight of the i-th main microphone; and N represents the total number of main microphones. S430. The initial weights of each main microphone are adjusted proportionally to obtain the weight coefficients of each main microphone; the formula for proportionally adjusting the initial weights is as follows: Where w[i]' represents the weight coefficient of the i-th main microphone.

2. The microphone array wind noise reduction optimization method according to claim 1, characterized in that, In step S100, the method for setting the reference microphone is as follows: Specify a microphone as the reference microphone in the microphone array, or set a separate microphone as the reference microphone.

3. The microphone array wind noise reduction optimization method according to claim 1, characterized in that: In step S100, the physical wind noise reduction device includes windproof down and / or windproof cover.

4. The microphone array wind noise reduction optimization method according to claim 1, characterized in that, After performing step S200, the following steps are also performed: S210. Store the audio signal data collected by the reference microphone and each main microphone as a time-domain signal sequence; S220. The time-domain signal sequence is subjected to denoising, filtering and amplification to obtain preprocessed audio data; S230. Extract time-domain or time-frequency domain features from the preprocessed audio data, and use the extracted features to characterize the signal acquired by the microphone.

5. The microphone array wind noise reduction optimization method according to any one of claims 1 to 4, characterized in that, In step S410, the formula for normalization is as follows: Where d_mse[i] represents the mean square error of the i-th master microphone; d_min represents the minimum mean square error among all master microphones; and d_max represents the maximum mean square error among all master microphones.

6. The microphone array wind noise reduction optimization method according to any one of claims 1 to 4, characterized in that, In step S500, the formula for optimizing the wind noise reduction of the signal acquired by the main microphone is as follows: y[i]=(1-w[i]')×x[i]+w[i]'×x_ref_processed Where y[i] represents the signal obtained after wind noise reduction optimization of the i-th main microphone; x[i] represents the signal collected by the i-th main microphone; and x_ref_processed represents the signal collected by the reference microphone.

7. A microphone array wind noise reduction optimization device, characterized in that: include A microphone array for acquiring ambient audio signals, the microphone array including a reference microphone and multiple main microphones, the reference microphone being equipped with a physical wind noise reduction device; The mean square error calculation module is used to calculate the mean square error of the signals acquired by each main microphone relative to the signals acquired by the reference microphone. The weight calculation module is used to calculate the weight coefficient of each main microphone based on the mean square error. as well as The wind noise reduction optimization module is used to optimize the wind noise reduction of the signal collected by the main microphone based on the weighting coefficient of the main microphone and the signal collected by the reference microphone. The method by which the weight calculation module calculates the weight coefficients of each main microphone based on the mean square error includes the following sub-steps: S410. Normalize the mean square error of each main microphone to obtain the normalized weight of each main microphone. S420. Calculate the initial weights of each main microphone based on their normalized weights. The formula for calculating the initial weights is as follows: Where i represents the sequence number of the main microphone; w[i] represents the initial weight of the i-th main microphone; d_norm[i] represents the normalized weight of the i-th main microphone; and N represents the total number of main microphones. S430. The initial weights of each main microphone are adjusted proportionally to obtain the weight coefficients of each main microphone; the formula for proportionally adjusting the initial weights is as follows: Where w[i]' represents the weight coefficient of the i-th main microphone.

8. The microphone array wind noise reduction optimization device according to claim 7, characterized in that: It also includes a preprocessing module, which is used to denoise, filter and amplify the audio signals collected by the reference microphone and each main microphone, and then extract their time domain features or time-frequency domain features, and use the extracted features to characterize the signals collected by the microphones.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is loaded and executed by the processor, it implements the microphone array wind noise reduction optimization method as described in any one of claims 1 to 6.

Images