Microphone array sound monitoring method and system based on multiple array elements

Abstract

The invention discloses a microphone array sound monitoring method and system based on multiple array elements, and relates to the field of equipment monitoring. Comprising the following steps: installing a microphone array at a monitoring device, and collecting a sound signal at the device by using the microphone array; the signal preprocessing module adopts a multi-channel technology to collect sound signals of the microphone array and preprocesses the sound signals; and the fault recognition model processes and recognizes the preprocessed signal based on a neural network algorithm, and judges the equipment fault state. According to the invention, fault detection is carried out on the substation equipment by using the sound signals, and the working voiceprint information of each piece of equipment is determined and is not influenced by a shielding object, so that the reliability and the accuracy of fault detection can be improved; in addition, compared with image recognition, voiceprint-based fault recognition has the advantages that the data processing amount is greatly reduced, and the computing power resource is saved.

Description

Technical Field

[0001] The present invention relates to the field of equipment monitoring, and in particular to a method and system for sound monitoring using a microphone array based on multiple array elements. Background Art

[0002] Substations are important hubs in the power transmission process, so early detection of faults in high-voltage equipment within the station is particularly important.

[0003] After searching, the Chinese patent publication number CN114265010A discloses a transformer fault sound source positioning method and system based on a microphone array. The method includes the following steps: S1. Acquire the sound signal x(t) at the measurement surface through the microphone array; S2. Use the improved multi-scale morphological filter to suppress the noise of the sound signal x(t) to obtain the filtered denoised signal, that is, the transformer equipment radiation noise signal in a pure state: S3. Perform time domain statistical analysis and spectrum analysis on the denoised signal, and compare it with the historical data in the healthy acoustic database: S4. Extract the frequency band of the abnormal denoised signal, search it in the fault acoustic database, and output the fault type and fault location corresponding to the frequency band to achieve the purpose of transformer equipment fault judgment and positioning.

[0004] The above patent has the following shortcomings: it only uses a filter to denoise the sound signal, which can only remove environmental noise but cannot remove the sound signal's own noise, thereby affecting the accuracy of subsequent processing. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a method and system for sound monitoring based on a microphone array with multiple elements.

[0006] The present invention adopts the following technical solution: a microphone array sound monitoring system based on multiple array elements, comprising: A microphone array is provided at the monitoring device and is used to collect sound in the air at the monitoring device; A signal preprocessing module receives the signal from the microphone array and preprocesses the signal; Fault identification model is established, and the preprocessed signal is input into the model for fault judgment.

[0007] A method for sound monitoring based on a microphone array with multiple elements comprises the following steps: S1: Install the microphone array at the monitoring device and use the microphone array to collect sound signals at the device; S2: The signal preprocessing module uses multi-channel technology to collect the sound signal of the microphone array and preprocess it; S3: The fault identification model processes and identifies the pre-processed signals based on the neural network algorithm to determine the equipment fault status.

[0008] Preferably: in step S2, the signal preprocessing comprises the following steps: S21: Improving signal-to-noise ratio using array positioning and directional enhancement algorithms; S22: Then locate the device position and shield the interference sound from other positions; S23: extracting voiceprint features from the sound signal, and classifying the signal according to the voiceprint features; S24: After enhancing or suppressing different types of signals, different channels are used for transmission.

[0009] Preferably: step S21 includes array positioning and directional enhancement steps.

[0010] Preferably: the array positioning comprises the following steps: A1: Signal preprocessing, preprocessing the original sound signal collected by the microphone array, including denoising, filtering, and gain adjustment; A2: Time difference estimation, using the generalized cross-correlation method to calculate the time difference between different microphones receiving the same sound source signal, and then correcting and compensating the time difference; A3: Sound source localization, using the least squares method or iterative method to calculate the position and direction of the sound source based on the geometric layout of the microphone array and the time difference information.

[0011] Preferably: the directional enhancement comprises the following steps: B1: Beamforming, constructing a signal vector and calculating the sound matrix. According to the sound source localization results, the direction and target of the beamforming are determined, the beamforming weight is calculated, and the signals received by the microphone array are weighted and summed to form a directional beam. B2: Adaptive filtering, using LMS filtering algorithm or RLS filtering algorithm to suppress noise and interference; B3: Signal reconstruction: The processed frequency domain signal is converted back to the time domain signal using the inverse short-time Fourier transform to perform de-jittering and smoothing on the reconstructed signal.

[0012] Preferably: step S23 comprises the following steps: S231: Time domain feature extraction, based on the time sequence characteristics of the sound, extract the time domain features, including pitch period, intonation, and energy; S232: frequency feature extraction; Perform Fourier transform on the sound signal to convert it into frequency domain representation, and then extract the amplitude spectrum information; First, the sound signal is Fourier transformed to obtain the spectrum, and then logarithmic transformation and inverse Fourier transformation are performed to obtain the cepstrum; Map the spectrum to the Mel frequency scale and then perform a logarithmic transformation; S233: Classification using Gaussian mixture general background model or joint factor analysis.

[0013] Preferably: in step S24, it includes the following steps: S241: Pre-establishing a model feature of the voiceprint of the device and forming a feature library; S242: After receiving the voice signal, compare the voiceprint signal of each channel with the voiceprint signal of the feature library to determine the channel corresponding to the voiceprint of the monitoring device; S243: Strengthen the channel corresponding to the voiceprint of the monitoring device, and suppress the other channels.

[0014] Preferably: in step S3, the method for fault diagnosis comprises the following steps: S31: further extracting a feature vector from the determined voiceprint of the device sound; S32: Calculate the similarity between the feature vector and the feature vector of the feature library; S33: Set a similarity threshold. If the similarity between the collected device voiceprint vector and the feature vector of the feature library is less than the threshold, a fault is determined.

[0015] Preferably: in step S32, the similarity calculation adopts any one of cosine similarity, Euclidean distance, Manhattan distance and Pearson correlation coefficient.

[0016] The present invention uses sound signals to detect faults in substation equipment. Since the voiceprint information of each device is certain and is not affected by obstructions, the reliability and accuracy of fault detection can be increased. In addition, the data processing volume of fault recognition based on voiceprint is greatly reduced compared to image recognition, saving computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of a multi-element microphone array sound monitoring method proposed by the present invention. DETAILED DESCRIPTION

[0018] The technical solution of the present invention is further described in detail below in conjunction with specific implementation methods.

[0019] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", and "set" should be understood in a broad sense, for example, they can be fixedly connected or set, or detachably connected or set, or integrally connected or set. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0020] Embodiment 1: A microphone array sound monitoring system based on multiple array elements, comprising: A microphone array is provided at the monitoring device and is used to collect sound in the air at the monitoring device; A signal preprocessing module receives the signal from the microphone array and preprocesses the signal; Fault identification model is established, and the preprocessed signal is input into the model for fault judgment.

[0021] The present invention uses sound signals to detect faults in substation equipment. Since the voiceprint information of each device is certain and is not affected by obstructions, the reliability and accuracy of fault detection can be increased. In addition, the data processing volume of fault recognition based on voiceprint is greatly reduced compared to image recognition, saving computing resources.

[0022] Embodiment 2: like Figure 1 As shown, a method for sound monitoring based on a microphone array with multiple elements is used to implement a sound monitoring system based on a microphone array with multiple elements in Example 1, comprising the following steps: S1: Install a microphone array at the monitoring device and use the microphone array to collect sound signals at the device; S2: The signal preprocessing module uses multi-channel technology to collect the sound signal of the microphone array and preprocess it; S3: The fault identification model processes and identifies the pre-processed signals based on the neural network algorithm to determine the equipment fault status.

[0023] In step S2, signal preprocessing includes the following steps: S21: Improving signal-to-noise ratio using array positioning and directional enhancement algorithms; S22: Then locate the device position and shield the interference sound from other positions; S23: extracting voiceprint features from the sound signal, and classifying the signal according to the voiceprint features; by classifying the voiceprint features, the sound signal of the device to be identified can be effectively distinguished from the other sound signals; S24: After enhancing and suppressing different types of signals, different channels are used for transmission; based on the classification, the sound transmission channel of the device to be identified is increased, and the sound signal of the device and the channel is suppressed, so as to effectively realize the noise reduction function.

[0024] Step S21 includes array positioning and directional enhancement steps.

[0025] The array positioning comprises the following steps: A1: Signal preprocessing, preprocessing the original sound signal collected by the microphone array, including denoising, filtering, and gain adjustment; A2: Time difference estimation, using the generalized cross-correlation method to calculate the time difference between different microphones receiving the same sound source signal, and then correcting and compensating the time difference; correcting and compensating the time difference of signals from different sound sources can increase the consistency of the output signals of multiple microphones; A3: Sound source localization, using the least squares method or iterative method to calculate the position and direction of the sound source based on the geometric layout of the microphone array and the time difference information.

[0026] The directional enhancement comprises the following steps: B1: Beamforming, constructing a signal vector and calculating the sound matrix. According to the sound source localization results, the direction and target of the beamforming are determined, the beamforming weight is calculated, and the signals received by the microphone array are weighted and summed to form a directional beam. B2: Adaptive filtering, using LMS filtering algorithm or RLS filtering algorithm to suppress noise and interference; B3: Signal reconstruction: The processed frequency domain signal is converted back to the time domain signal using the inverse short-time Fourier transform to perform de-jittering and smoothing on the reconstructed signal.

[0027] Step S23 includes the following steps: S231: Time domain feature extraction, based on the time sequence characteristics of the sound, extract the time domain features, including pitch period, intonation, and energy; S232: frequency feature extraction; Perform Fourier transform on the sound signal to convert it into frequency domain representation, and then extract the amplitude spectrum information; First, the sound signal is Fourier transformed to obtain the spectrum, and then logarithmic transformation and inverse Fourier transformation are performed to obtain the cepstrum; Map the spectrum to the Mel frequency scale and then perform a logarithmic transformation; S233: Classification using Gaussian mixture general background model or joint factor analysis.

[0028] In step S24, it includes the following steps: S241: Pre-establishing a model feature of the voiceprint of the device and forming a feature library; S242: After receiving the voice signal, the voiceprint signal of each channel is compared with the voiceprint signal of the feature library to determine the channel corresponding to the voiceprint of the monitoring device. By comparing the voiceprint library signal, the sound signal of the device under test can be accurately distinguished from the other sound signals; S243: Strengthen the channel corresponding to the voiceprint of the monitoring device, and suppress the other channels, thereby achieving noise suppression.

[0029] In step S3, the method for fault diagnosis includes the following steps: S31: further extracting a feature vector from the determined voiceprint of the device sound; S32: Calculate the similarity between the feature vector and the feature vector of the feature library; S33: Set a similarity threshold. If the similarity between the collected device voiceprint vector and the feature vector of the feature library is less than the threshold, a fault is determined. The similarity calculation of the feature vector is based on the probability principle and can reduce the misjudgment of faults.

[0030] In step S32, the similarity calculation adopts any one of cosine similarity, Euclidean distance, Manhattan distance, and Pearson correlation coefficient.

[0031] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A microphone array sound monitoring system based on multiple array elements, characterized in that: include: A microphone array is provided at the monitoring device and is used to collect sound in the air at the monitoring device; A signal preprocessing module receives the signal from the microphone array and preprocesses the signal; Fault identification model is established, and the preprocessed signal is input into the model for fault judgment.

2. A method for sound monitoring based on a microphone array with multiple elements, used to implement a sound monitoring system based on a microphone array with multiple elements as claimed in claim 1, characterized in that: The following steps are involved: S1: Install the microphone array at the monitoring device and use the microphone array to collect sound signals at the device; S2: The signal preprocessing module uses multi-channel technology to collect the sound signal of the microphone array and preprocess it; S3: The fault identification model processes and identifies the pre-processed signals based on the neural network algorithm to determine the equipment fault status.

3. The method for sound monitoring based on a multi-element microphone array according to claim 2, characterized in that: In step S2, signal preprocessing includes the following steps: S21: Improving signal-to-noise ratio using array positioning and directional enhancement algorithms; S22: Then locate the device position and shield the interference sound from other positions; S23: extracting voiceprint features from the sound signal, and classifying the signal according to the voiceprint features; S24: After enhancing and suppressing different types of signals, different channels are used for transmission.

4. The method for sound monitoring based on a multi-element microphone array according to claim 3, characterized in that: Step S21 includes array positioning and directional enhancement steps.

5. The method for sound monitoring based on a multi-element microphone array according to claim 4, characterized in that: The array positioning comprises the following steps: A1: Signal preprocessing, preprocessing the original sound signal collected by the microphone array, including denoising, filtering, and gain adjustment; A2: Time difference estimation, using the generalized cross-correlation method to calculate the time difference between different microphones receiving the same sound source signal, and then correcting and compensating the time difference; A3: Sound source localization, using the least squares method or iterative method to calculate the position and direction of the sound source based on the geometric layout of the microphone array and the time difference information.

6. The method for sound monitoring based on a multi-element microphone array according to claim 4, characterized in that: The directional enhancement comprises the following steps: B1: Beamforming, constructing a signal vector and calculating the sound matrix. According to the sound source localization results, the direction and target of the beamforming are determined, the beamforming weight is calculated, and the signals received by the microphone array are weighted and summed to form a directional beam. B2: Adaptive filtering, using LMS filtering algorithm or RLS filtering algorithm to suppress noise and interference; B3: Signal reconstruction: The processed frequency domain signal is converted back to the time domain signal using the inverse short-time Fourier transform to perform de-jittering and smoothing on the reconstructed signal.

7. The method for sound monitoring based on a multi-element microphone array according to claim 4, characterized in that: Step S23 includes the following steps: S231: Time domain feature extraction, based on the time sequence characteristics of the sound, extract the time domain features, including pitch period, intonation, and energy; S232: frequency feature extraction; Perform Fourier transform on the sound signal to convert it into frequency domain representation, and then extract the amplitude spectrum information; First, the sound signal is Fourier transformed to obtain the spectrum, and then logarithmic transformation and inverse Fourier transformation are performed to obtain the cepstrum; Map the spectrum to the Mel frequency scale and then perform a logarithmic transformation; S233: Classification using Gaussian mixture general background model or joint factor analysis.

8. The method for sound monitoring based on a multi-element microphone array according to claim 4, characterized in that: In step S24, it includes the following steps: S241: Pre-establishing a model feature of the voiceprint of the device and forming a feature library; S242: After receiving the voice signal, compare the voiceprint signal of each channel with the voiceprint signal of the feature library to determine the channel corresponding to the voiceprint of the monitoring device; S243: Strengthen the channel corresponding to the voiceprint of the monitoring device, and suppress the other channels.

9. The method for sound monitoring based on a multi-element microphone array according to claim 2, characterized in that: In step S3, the method for fault diagnosis includes the following steps: S31: further extracting a feature vector from the determined voiceprint of the device sound; S32: Calculate the similarity between the feature vector and the feature vector of the feature library; S33: Set a similarity threshold. If the similarity between the collected device voiceprint vector and the feature vector of the feature library is less than the threshold, a fault is determined.

10. The method for sound monitoring based on a multi-element microphone array according to claim 9, characterized in that: In step S32, the similarity calculation adopts any one of cosine similarity, Euclidean distance, Manhattan distance, and Pearson correlation coefficient.

Images