A distributed ultra-weak grating microphone array sound source positioning method and system

Abstract

This invention relates to the field of acoustic logging technology and discloses a method and system for locating sound sources using a distributed ultra-weak grating microphone array. The method involves acquiring acoustic signals from a target area using microphones at different locations, preprocessing the acquired signals, and extracting phase amplitude and time delay features from the preprocessed signal data. Based on a pre-trained model, the phase amplitude and time delay features are matched to determine the location and distance of the sound source. By analyzing the phase amplitude and signal delay between the sound source and different microphones, the distance and angle of the sound source relative to a reference microphone can be accurately determined. This method utilizes the time delay and amplitude features of the microphone array, improving the accuracy of sound source localization. It also makes the sound source signal more prominent in strong noise backgrounds, facilitating identification and analysis.

Description

Technical Field

[0001] This invention relates to the field of acoustic logging technology, specifically to a method and system for locating sound sources using a distributed ultra-weak grating microphone array. Background Technology

[0002] Microphone array sound source localization technology captures sound signals simultaneously through multiple microphones, accurately measuring the location and intensity of sound sources. It has been widely applied in various fields such as smart homes, security monitoring, and drone navigation. Currently, this technology is gradually entering the oil industry, becoming an emerging noise logging method. It can collect vibration information from formations, which can be used to assess wellbore integrity, detect oil and gas reservoirs, identify formation fractures and faults, and evaluate oil and gas well production and flow characteristics, thereby optimizing production processes and improving the efficiency of oil and gas exploration and recovery.

[0003] One existing technology uses a broadband light source to radiate broadband light, which is then split into three beams in a fiber optic coupler. Each beam is reflected by a quartz glass probe, and two beams return to the coupler and interfere. However, this method can only create a few microphones, and the detection mechanism is large and cannot be lowered into the well, making it unsuitable for well logging operations. Another method can provide sound source localization in the XYZ coordinate system on the ground, but it is not suitable for downhole polar coordinate space localization. Therefore, there is currently no suitable sound acquisition method for the well logging industry. In summary, existing microphone array sound source localization technology detects noise through acoustic array probes, which makes it difficult to design downhole instruments. Furthermore, due to the limited number of acoustic probes, it cannot simultaneously detect noise throughout the entire well, limiting testing to point measurements and significantly restricting application scenarios. In contrast, while distributed acoustic sensing (DAS) systems can be used as microphone arrays, their weak backscattered Rayleigh light signal results in a low signal-to-noise ratio after photoelectric conversion, affecting signal acquisition quality and leading to inaccurate sound source localization and intensity measurement errors. Summary of the Invention

[0004] Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a method and system for locating sound sources using a distributed ultra-weak grating microphone array. This method has the advantage of accurately determining the location and intensity of sound sources in underground mines, thus solving the problem of the inability to accurately determine the location and intensity of sound sources in existing technologies.

[0005] Technical solution To achieve the above objectives, the present invention provides the following technical solution: A method for localizing sound sources using a distributed ultra-weak grating microphone array includes the following steps: S1. Acquire sound signals from the target area using microphones at different locations, and perform preprocessing operations on the acquired sound signals; S2. Perform feature extraction on the preprocessed acoustic signal data to obtain phase amplitude features and time delay features; S3. Based on the pre-trained model, the location and distance of the sound source are obtained by performing feature recognition and matching of phase amplitude features and time delay features.

[0006] Preferably, the sound signal is acquired through a UW-FBG microphone, and the sound signal is acquired through M UW-FBG microphones and converted into M-channel digital signal data.

[0007] Preferably, the preprocessing operation of the acquired acoustic signal includes: using Butterworth filtering to remove the DC component from the original signal, and using LMS adaptive filtering to perform noise reduction processing on the acoustic signal after removing the DC component.

[0008] Preferably, the pre-trained model adopts an M-ary linear localization model.

[0009] Preferably, the pre-trained model is obtained by training the amplitude features and time delay features extracted from the pre-trained acoustic signal.

[0010] Preferably, the amplitude feature extracted from the pre-trained acoustic signal is a log-Mel spectrum feature; the time delay feature extracted from the pre-trained acoustic signal is a SCOT / PHAT jointly weighted generalized cross-correlation feature.

[0011] A distributed ultra-weak grating microphone array sound source localization system includes a sound signal acquisition and preprocessing module, a signal feature extraction module, and a sound source localization module; The sound signal acquisition and preprocessing module acquires sound signals from the target area through microphones at different locations and performs preprocessing operations on the acquired sound signals. The signal feature extraction module extracts phase amplitude features and time delay features from the preprocessed acoustic signal data. The sound source localization module uses a pre-trained model to perform feature recognition and matching of phase amplitude features and time delay features to obtain the location and distance of the sound source.

[0012] Preferably, the acoustic signal acquisition module uses UW-FBG microphones to acquire acoustic signals from the target area. The UW-FBG microphones are constructed from single-mode, high-bending-fatigue-resistance ultra-weak optical fibers and thin-walled polycarbonate cylinders, with a length of 100mm, an outer diameter of 20mm, and a wall thickness of 0.5mm. Three UW-FBG microphones (…) are then used. Signals are acquired by arranging the components in series at equal intervals.

[0013] Preferably: in the acoustic signal acquisition and preprocessing module, the acoustic signal... After Butterworth filters out the low-frequency DC signal, the result is... Then, the LMS algorithm is used for noise reduction to obtain... .

[0014] Preferably, the Butterworth filter principle formula is:

[0015] in, The cutoff frequency, The passband edge frequency, This indicates the order of the Butterworth filter.

[0016] Preferably, the digital signal still contains complex noise after being filtered by the Butterworth filter formula. The signal is then further processed using the LMS algorithm. The output signal is calculated by using the input signal and weight coefficients. The adaptive adjustment is to square the error response signal obtained from the reference response and the output response and obtain its mean square minimum value. Then, the weight value is updated through logical judgment, thereby achieving the best filtering operation for the signal.

[0017] Compared with the prior art, the present invention provides a method and system for localizing sound sources using a distributed ultra-weak grating microphone array, which has the following beneficial effects: This invention discloses a distributed ultra-weak grating microphone array sound source localization method. It collects sound signals from a target area using microphones at different locations, preprocesses the acquired sound signals, and extracts phase amplitude and time delay features from the preprocessed sound signal data. Based on a pre-trained model, it performs feature recognition and matching on the phase amplitude and time delay features to obtain the location and distance of the sound source. By analyzing the phase amplitude and signal delay between the sound source and different microphones, it accurately infers the distance and angle of the sound source relative to a reference microphone. This method utilizes the time delay and amplitude features of the microphone array to improve the accuracy of sound source localization.

[0018] This invention utilizes a distributed ultra-weak grating microphone array to collect acoustic signals from the target area using M (M≧3) UW-FBG microphones, converting them into digital signals. This enhances the amplitude of the overall downhole noise signal, providing diverse data support for whole-well noise analysis. Furthermore, signal preprocessing techniques, including Butterworth filter removal of DC components and LMS algorithm for denoising, improve the signal-to-noise ratio, making the sound source signal stand out more against a strong noise background, facilitating identification and analysis.

[0019] This invention establishes an M-ary linear localization model and extracts features from the preprocessed signal, including log-Mel spectrum features and SCOT / PHAT jointly weighted generalized cross-correlation features. These features are then input into a CRNN network model for training and testing. By utilizing the learning capabilities of the neural network, the accuracy and intelligence of sound source localization can be further improved. While ensuring localization accuracy, the model can be simplified and costs controlled. Furthermore, by optimizing the layout of the microphone array and the signal processing algorithm, the system complexity and cost can be reduced. Attached Figure Description

[0020] Figure 1 This is a flowchart of the distributed ultra-weak grating microphone array sound source localization method of the present invention; Figure 2 This is a diagram of the sound source localization model of the distributed ultra-weak grating microphone array of the present invention; Figure 3 This is a simulation diagram of the time delay of the sound source localization model of the present invention; Figure 4 This is a flowchart of the sound source localization network of the present invention. Detailed Implementation

[0021] 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.

[0022] Please see Figure 1 - Figure 4 A method for localizing sound sources using a distributed ultra-weak grating microphone array includes the following steps: S1. Acquire sound signals from the target area using microphones at different locations, and perform preprocessing operations on the acquired sound signals in sequence; S2. Preprocess the acquired acoustic signal: Use a Butterworth filter to remove the DC component from the signal, and then use the LMS algorithm for noise reduction; S3. Establish an M-element linear positioning model and analyze the positioning principle. By obtaining the phase amplitude and the delay between the signal, the distance and angle of the sound source to the reference microphone can be deduced. S4. Extract features from the preprocessed signal and input it into a neural network for learning to achieve localization. Extract the acoustic signal based on the amplitude characteristics of the acoustic signal. S5. Introduce the CRNN network model, train and test it to determine the location of the sound source and estimate the distance.

[0023] The sound signal is acquired through UW-FBG microphones, and the sound signal is acquired through M UW-FBG microphones. The acquired sound signal is then converted into M-channel digital signal data through acousto-optic modulation, photoelectric conversion, and analog-to-digital conversion.

[0024] The signal preprocessing methods for acoustic signals include: processing with Butterworth filtering and LMS adaptive filtering. Butterworth filtering can remove the DC component from the original acoustic signal, while LMS adaptive filtering can improve the signal-to-noise ratio. Further noise reduction processing is then performed on the acoustic signal after removing the DC component.

[0025] The M-element linear localization model includes: determining the location and distance of the sound source by establishing a linear array localization model and analyzing the phase amplitude and time delay difference characteristics of the signal.

[0026] The sound source localization method extracts corresponding features from the phase amplitude characteristics and time delay difference characteristics of the array signal, and inputs the acquired features into the CRNN network for training and testing. Among them, the corresponding features extracted from the phase amplitude characteristics and time delay difference characteristics are called composite features. The composite features are the log-Mel spectrum features representing the amplitude characteristics and the SCOT / PHAT joint weighted generalized cross-correlation features representing the time delay characteristics. The CRNN neural network is introduced for learning, and the accurate localization of the target sound source is achieved.

[0027] like Figure 1 The diagram shown is a flowchart of the distributed ultra-weak grating microphone array sound source localization method of the present invention, which specifically includes the following steps: Signal acquisition: A system is constructed using M (M≧3) UW-FBG microphones to collect acoustic signals from the target area. The collected acoustic signals are then processed through acousto-optic modulation, photoelectric conversion, and analog-to-digital conversion to obtain digital signals. (i=1) 2,m).

[0028] Signal preprocessing: A Butterworth filter is used to remove the DC component from the signal. Then, the LMS algorithm is used to process the acquired signal. Denoising is performed to improve the signal-to-noise ratio. ; Positioning Model: Establish an M-element linear positioning model and analyze its positioning principle. This analysis yields... The phase amplitude and the delay between the signal can reflect the distance and angle of the sound source to the reference microphone.

[0029] For the preprocessed signal Log-Mel spectrum features were extracted to obtain Extracting the SCOT / PHAT jointly weighted generalized cross-correlation features yields... We introduced a CRNN network model for training and testing to achieve sound source localization.

[0030] This invention provides a distributed ultra-weak grating microphone array sound source localization system, characterized in that it includes a sound signal acquisition and preprocessing module, a signal feature extraction module, and a sound source localization module; The sound signal acquisition and preprocessing module acquires sound signals from the target area through microphones at different locations and performs preprocessing operations on the acquired sound signals. The signal feature extraction module extracts phase amplitude features and time delay features from the preprocessed acoustic signal data. The sound source localization module uses a pre-trained model to perform feature recognition and matching of phase amplitude features and time delay features to obtain the location and distance of the sound source.

[0031] The acoustic signal acquisition and preprocessing module uses UW-FBG microphones to acquire acoustic signals from the target area. The UW-FBG microphones are constructed from single-mode, high-bending-fatigue-resistance ultra-weak optical fibers and thin-walled polycarbonate cylinders, with a length of 100mm, an outer diameter of 20mm, and a wall thickness of 0.5mm. Three UW-FBG microphones (…) Signals are acquired by arranging the components in series at equal intervals.

[0032] In the acoustic signal acquisition and preprocessing module, digital signals After Butterworth filters out the low-frequency DC signal, the result is... Then, the LMS algorithm is used for noise reduction to obtain... .

[0033] The formula for Butterworth filtering is:

[0034] in, The cutoff frequency, The passband edge frequency, This indicates the order of the Butterworth filter. The Butterworth filter ensures that the frequency response curve is as flat as possible in the passband, while gradually decreasing to zero in the stopband.

[0035] Even after being filtered by the Butterworth filter, digital signals still contain complex noise. The LMS algorithm is then used to further process the signal. The output signal is calculated by using the input signal and weight coefficients. The adaptive adjustment involves squaring the error response signal obtained from the reference response and the output response, finding its mean square minimum, and then updating the weight values ​​through logical judgment, thereby achieving the best filtering operation for the signal.

[0036] Suppose that the filter bank has If there are several adaptive filters, then the weight coefficients of each filter are: The filter weights after passing through the filter bank for:

[0037] Output signal with weight coefficients and input signal The relationship can be represented as:

[0038] Output signal With reference signal Subtraction yields the error signal. for:

[0039] Iteratively update the weight vector using the squared error method:

[0040] in, The iteration step size, The weighting coefficients are continuously adjusted based on the error between the filtered signal and the original signal to achieve the optimal filtering operation for the signal.

[0041] A linear positioning model of M elements is established and simulation analysis is performed. A sound source localization model is established using the reference center, such as... Figure 2 As shown, let the microphone spacing be... The direction of the sound source and The azimuth angle between them is The pitch angle is ,in, for If the sound source is in a two-dimensional model, then it is a three-dimensional model; otherwise, it is a three-dimensional model. (Sound source distance) distance for and Then the distance of the sound source distance for:

[0042] Because the distance from the sound source to the microphone varies, the attenuation of the sound source varies within the strata, and the microphone's response to sound signals from different directions also varies, amplitude differences arise. The sound source propagation time also varies, resulting in time delay differences. Amplitude differences are directly reflected in the amplitude characteristics through the amplitude of the received signal, while time delay differences are more complex; therefore, a mathematical model needs to be constructed for analysis. The time delay difference between them ( For example, then:

[0043] The sound source localization model time delay difference simulation is established using formula (3), as follows: Figure 3 As shown, the model parameters are set as follows: , Angular interval is ; The distance interval is 1. From Figure 3 As can be seen, different angles and distances will lead to... The differences in these factors lead to differences in time delay characteristics.

[0044] Log-Mel spectrum features were extracted based on the amplitude characteristics of the acoustic signal, and SCOT / PHAT joint weighted generalized cross-correlation features were extracted based on the time delay characteristics of the acoustic signal. A CRNN was constructed for feature learning, and sound source localization was achieved.

[0045] The positioning process is as follows Figure 4 The process is as follows: S1: Extracting Log-Mel Spectrum Features. The denoised acoustic signal is framed and windowed to become a short-time signal, which is then subjected to Fourier transform and Mel filtering to obtain the log-Mel spectrum signal.

[0046] (1) Framing and windowing the original signal:

[0047] Where n represents the sample number in the window ( N is the window length. Represents the original time-domain signal. Indicates the first Window weighting coefficients for each sample, This represents the time-domain signal after windowing.

[0048] (2) Fast Fourier Transform:

[0049] in, It is the time-frequency signal after FFT transformation: (3) The power spectrum of the time-frequency signal obtained after FFT is calculated by Mel filtering, and then passed through the Mel filter bank and logarithmic operation is performed, as follows:

[0050] Where M represents the number of filters. Here is the filter transfer function:

[0051] S2: Extract the generalized cross-correlation features of the SCOT / PHAT joint weighting. For multi-channel signals, there will be a time delay between each pair of signals. Based on the cross-correlation spectrum formula, the relationship between the delay and the spectral power is expressed as follows: (1) Fourier Transform. (This is applied to the pairs of elements.) and Received signal and Perform a Fourier transform to obtain and .

[0052] (2) Calculate the cross power spectrum:

[0053] in, for Conjugate.

[0054] (3) Calculate the generalized cross-correlation spectrum:

[0055] in, For array element and The time delay difference of the received signal, This is the weighting factor.

[0056] (4) Calculate the SCOT / PHAT joint weighted generalized cross-correlation spectrum: SCOT / PHAT joint weighting factor The calculation formula is as follows:

[0057] The CRNN is used to train the feature training sample set to obtain the corresponding acoustic model. The CRNN performs feature convolution calculation on the input features through 4 layers of 2D CNN convolutional layers. The convolution kernel size is 3×3, the activation function is ReLU, and max pooling is performed after each convolution. After completion, the result is fed into a bidirectional GRU layer to obtain information in the time dimension. Finally, the result is output through a fully connected layer.

[0058] The test sample set is input into the acoustic model to determine the direction of the sound source and estimate the distance.

[0059] This invention employs ultra-weak fiber grating (UW-FBG) technology as a novel approach in fiber optic sensing. By irradiating the fiber with ultraviolet light or femtosecond laser, a permanent periodic change in its refractive index is caused, forming thousands of gratings on a single fiber. The optical signals reflected by these gratings are stable, and each grating functions as a microphone. This technology enables large-scale, high-precision, and highly consistent acoustic wave detection. Combining UW-FBG arrays and microphone array sound source localization technology yields a consistent and stable downhole noise acquisition scheme, accurately determining the location and intensity of downhole sound sources, thus meeting the real-time radial formation monitoring needs of oilfield engineering.

[0060] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for sound source localization using a distributed ultra-weak grating microphone array, characterized in that, Includes the following steps: S1. Acquire sound signals from the target area using microphones at different locations, and perform preprocessing operations on the acquired sound signals; S2. Perform feature extraction on the preprocessed acoustic signal data to obtain phase amplitude features and time delay features; S3. Based on the pre-trained model, the location and distance of the sound source are obtained by performing feature recognition and matching of phase amplitude features and time delay features.

2. The method for localizing a sound source using a distributed ultra-weak grating microphone array according to claim 1, characterized in that, The sound signal is acquired through a UW-FBG microphone, and the sound signals are acquired through M UW-FBG microphones and converted into M-channel digital signal data.

3. The method for localizing a sound source using a distributed ultra-weak grating microphone array according to claim 1, characterized in that, The preprocessing operation of the acquired acoustic signal includes: using Butterworth filtering to remove the DC component from the original signal, and using LMS adaptive filtering to denoise the acoustic signal after removing the DC component.

4. The method for localizing a sound source using a distributed ultra-weak grating microphone array according to claim 1, characterized in that, The pre-trained model uses an M-ary linear localization model.

5. The method for localizing a sound source using a distributed ultra-weak grating microphone array according to claim 4, characterized in that, The pre-trained model is obtained by training the amplitude and time delay features extracted from the pre-trained acoustic signal.

6. The method for localizing a sound source using a distributed ultra-weak grating microphone array according to claim 5, characterized in that, The amplitude features extracted from the pre-trained acoustic signal are log-Mel spectrum features; The time delay features extracted from the pre-trained acoustic signal are SCOT / PHAT jointly weighted generalized cross-correlation features.

7. A distributed ultra-weak grating microphone array sound source localization system, characterized in that, It includes a sound signal acquisition and preprocessing module, a signal feature extraction module, and a sound source localization module; The sound signal acquisition and preprocessing module acquires sound signals from the target area through microphones at different locations and performs preprocessing operations on the acquired sound signals. The signal feature extraction module extracts phase amplitude features and time delay features from the preprocessed acoustic signal data. The sound source localization module uses a pre-trained model to perform feature recognition and matching of phase amplitude features and time delay features to obtain the location and distance of the sound source.

8. The distributed ultralow-power grating microphone array sound source positioning system according to claim 7, characterized in that: The acoustic signal acquisition module uses UW-FBG microphones to acquire acoustic signals from the target area. The UW-FBG microphones are constructed from single-mode, high-bending-fatigue-resistance ultra-weak optical fibers and thin-walled polycarbonate cylinders, with a length of 100mm, an outer diameter of 20mm, and a wall thickness of 0.5mm. Three UW-FBG microphones (…) are then used. Signals are acquired by arranging the components in series at equal intervals.

9. A distributed ultra-weak grating microphone array sound source localization system according to claim 6, characterized in that: In the acoustic signal acquisition and preprocessing module, the acoustic signal After Butterworth filters out the low-frequency DC signal, the result is... Then, the LMS algorithm is used for noise reduction to obtain... .

10. A distributed ultra-weak grating microphone array sound source localization system according to claim 9, characterized in that: The formula for the Butterworth filter principle is as follows: in, The cutoff frequency, The passband edge frequency, This indicates the order of the Butterworth filter.

Images