Noise-containing source extraction and positioning method and system

Abstract

The invention discloses a noise source extraction and positioning method and system, and the method comprises the steps: obtaining a target mixed sound signal collected by a target unmanned plane, converting the target mixed sound signal into a frequency domain signal, and obtaining an amplitude spectrum matrix; performing matrix decomposition on the amplitude spectrum matrix, and obtaining a cross-correlation function in combination with a frequency domain weighting function; according to the cross-correlation function, determining the time difference of propagation of the target noise-containing source to the target unmanned aerial vehicle sound source acquisition device, and according to the coordinate relationship, determining the spatial position of the target noise-containing source; and filtering the amplitude spectrum matrix according to the cross-correlation function in combination with a binary mask matrix based on a preset interference judgment rule to obtain a noise-filtered amplitude spectrum matrix. Through the technical means of a frequency domain weighting function, a matrix decomposition technology, a binary mask matrix based on a preset interference judgment rule and the like, the positioning precision is effectively improved, the calculation complexity is reduced, the anti-interference capability is enhanced, rapid positioning is realized, and the method has a wide application range.

Description

Technical Field

[0001] The present invention relates to the technical field of noise source extraction and positioning, and in particular to a noise source extraction and positioning method and system. Background Art

[0002] In modern power systems, transformers are key equipment, and their operating status is directly related to the stability and reliability of the power system. However, transformers produce certain sounds during operation, which not only affect the environment, but may also indicate equipment failure or abnormality. Therefore, by timely identifying and locating the source of transformer sound, existing problems can be discovered and maintained in a timely manner, thereby avoiding power system shutdown or damage caused by equipment failure.

[0003] With the rapid development of drone technology, drone microphone arrays have shown great application potential in the fields of environmental noise monitoring and sound source localization. Drones have the advantages of high flexibility, wide coverage, and easy operation, which enables them to be flexibly deployed in complex and changing environments. Compared with traditional fixed noise monitoring equipment, drones can perform multi-directional noise monitoring of transformers at different heights and angles, thereby improving the comprehensiveness and accuracy of detection.

[0004] However, despite the many advantages of drone microphone arrays in noise monitoring, they still face challenges in practical applications. Drones generate significant self-noise during flight, which will mix with the transformer sound, increasing the complexity of signal processing. At the same time, since drones usually operate in outdoor environments, the sound signals collected by their microphone arrays may also be interfered by environmental noise such as wind and bird calls, which further increases the difficulty of locating the transformer sound source. Therefore, it is necessary to first use advanced signal processing technology to separate and extract the target sound signal, thereby effectively reducing the interference of drone self-noise and environmental noise on transformer sound monitoring. Then, the spatial information of the microphone array is used to locate the target sound source based on the differences in the sounds collected by microphones at different locations. Summary of the invention

[0005] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0006] In view of the above existing problems, the present invention is proposed.

[0007] Therefore, the present invention provides a noise source extraction and positioning method and system, which can solve the problems mentioned in the background technology.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0009] In a first aspect, the present invention provides a method for extracting and locating a noise source, comprising:

[0010] Acquire the target mixed sound signal collected by the left and right sound source collection devices of the target UAV, convert the target mixed sound signal into a frequency domain signal, and obtain an amplitude spectrum matrix, wherein the amplitude spectrum matrix includes a left amplitude spectrum matrix and a right amplitude spectrum matrix;

[0011] Performing matrix decomposition on the amplitude spectrum matrix to obtain a frequency domain dictionary matrix and a time domain coefficient matrix, and obtaining a cross-correlation function in combination with a frequency domain weighting function;

[0012] Determine the time difference of the target noise source propagating to the target UAV sound source collection device according to the cross-correlation function, and determine the spatial position of the target noise source according to the coordinate relationship;

[0013] At the same time, according to the cross-correlation function, combined with a binary mask matrix based on a preset interference judgment rule, the amplitude spectrum matrix is ​​filtered to obtain the amplitude spectrum matrix after the noise is filtered out.

[0014] As a preferred solution of the noise source extraction and positioning method described in the present invention, it also includes: performing an inverse short-time Fourier transform on the amplitude spectrum matrix after the noise is filtered out, and then superimposing the time domain signals to obtain the extracted and denoised time domain target sound.

[0015] As a preferred solution of the noise source extraction and positioning method of the present invention, wherein: the matrix decomposition of the amplitude spectrum matrix includes:

[0016] Decomposing the amplitude spectrum matrix into a frequency domain dictionary matrix and a time domain coefficient matrix;

[0017] The frequency domain dictionary matrix is ​​obtained by the following expression:

[0018]

[0019] The time domain coefficient matrix is ​​obtained by the following expression:

[0020]

[0021] Among them, ⊙ represents the multiplication of the corresponding elements of the matrix, α is the sparsity control parameter, V0 represents the amplitude spectrum matrix, that is, the left amplitude spectrum matrix or the right amplitude spectrum matrix, H n Represents the time domain coefficient matrix of the nth magnitude spectrum matrix, W n The frequency domain dictionary matrix representing the nth magnitude spectrum matrix, Hn T represents the transpose of the time domain coefficient matrix of the nth magnitude spectrum matrix, W n+1 The frequency domain dictionary matrix representing the n+1th magnitude spectrum matrix, H n+1 The time domain coefficient matrix of the n+1th magnitude spectrum matrix.

[0022] As a preferred solution of the noise source extraction and positioning method of the present invention, the method of obtaining the cross-correlation function by combining the frequency domain weighted function includes:

[0023] The signal is filtered using a frequency domain weighting function to highlight the spectrum components of the response signal part and suppress the spectrum of the noise part;

[0024]

[0025] Among them, W f Represents the row vector corresponding to frequency f in the dictionary matrix W, V l represents the left amplitude spectrum matrix, V r represents the right amplitude spectrum matrix, W represents the frequency domain dictionary matrix of the amplitude spectrum matrix, and the frequency domain dictionary matrix includes the left frequency domain dictionary matrix and the right frequency domain dictionary matrix.

[0026] As a preferred solution of the noise source extraction and positioning method of the present invention, the cross-correlation function includes:

[0027] According to the left frequency domain dictionary matrix and the right frequency domain dictionary matrix, in combination with the frequency domain weighting function, the left frequency domain weighting function and the right frequency domain weighting function of the target mixed sound signal collected by the left sound source collection device are calculated, and the mean of the left frequency domain weighting function and the right frequency domain weighting function is obtained;

[0028] The cross-correlation function is expressed as:

[0029]

[0030] Where * represents the complex conjugate of the matrix, and the cross-correlation function reflects the function of the sound arrival time difference τ corresponding to each dictionary atom d at each time t, ψ f represents the mean of the left frequency domain weighted function and the right frequency domain weighted function, ⊙ represents the multiplication of the corresponding elements of the matrix, Re represents the real part, e j2πfτ Indicates plural units.

[0031] As a preferred solution of the noise source extraction and positioning method of the present invention, wherein: the time difference of the target noise source propagating to the target UAV sound source collection device according to the cross-correlation function, and the spatial position of the target noise source according to the coordinate relationship includes:

[0032] The maximum value of the cross-correlation function in each time period is recorded as the first time difference, where the maximum value is the time difference between the target sound propagating to the left and right sound source collection devices;

[0033] The spatial position of the target noise source is obtained based on the spatial coordinates of the left and right sound source collection devices and the target sound arrival time difference.

[0034] As a preferred solution of the noise source extraction and positioning method of the present invention, the binary mask matrix based on the preset interference judgment rule includes:

[0035] When the base spectrum is not considered, the frequency domain weighting function is expressed as follows:

[0036]

[0037] Recalculate the cross-correlation function according to the frequency domain weighting function without considering the base spectrum, and the arrival time difference corresponding to the maximum value of the cross-correlation function, which is recorded as the second time difference;

[0038] If the difference between the first time difference and the second time difference is within a preset threshold, the base spectrum is considered to belong to the target signal, otherwise it is related to interference;

[0039] Use the binary mask matrix M dt This association is shown as follows:

[0040]

[0041] Among them, τ dt represents the first time difference, τ t represents the second time difference, ε represents the preset threshold;

[0042] Through the binary mask matrix M dt Get the amplitude spectrum matrix V after filtering out noise ft ', expressed as:

[0043] V ft '=W fd (H dt ⊙M dt )

[0044] Among them, W fd Represents the frequency domain dictionary matrix, including the left frequency domain dictionary matrix W l，fd , right frequency domain dictionary matrix W r，fd ;H dt Represents the time domain coefficient matrix, including the left time domain coefficient matrix H l，dt , right time domain coefficient matrix H r，dt , d is the dictionary atom, representing the base spectrum.

[0045] In a second aspect, the present invention provides a noise source extraction and positioning system, comprising:

[0046] An amplitude spectrum acquisition module is used to acquire the target mixed sound signal collected by the left and right sound source acquisition devices of the target UAV, convert the target mixed sound signal into a frequency domain signal, and obtain an amplitude spectrum matrix, wherein the amplitude spectrum matrix includes a left amplitude spectrum matrix and a right amplitude spectrum matrix;

[0047] A cross-correlation function calculation module, used to perform matrix decomposition on the amplitude spectrum matrix to obtain a frequency domain dictionary matrix and a time domain coefficient matrix, and obtain a cross-correlation function in combination with a frequency domain weighting function;

[0048] A positioning module, used to determine the time difference of the target noise source propagating to the target UAV sound source collection device according to the cross-correlation function, and determine the spatial position of the target noise source according to the coordinate relationship;

[0049] The extraction module is used to filter the amplitude spectrum matrix according to the cross-correlation function and the binary mask matrix based on the preset interference judgment rule to obtain the amplitude spectrum matrix after the noise is filtered out.

[0050] In a third aspect, the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above-mentioned method when executing the computer program.

[0051] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the method described above when executed by a processor.

[0052] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention proposes a noise source extraction and positioning method and system, obtains the target mixed sound signal collected by the left and right sound source collection devices of the target UAV, converts the target mixed sound signal into a frequency domain signal, and obtains an amplitude spectrum matrix, the amplitude spectrum matrix includes a left amplitude spectrum matrix and a right amplitude spectrum matrix; the amplitude spectrum matrix is ​​matrix decomposed to obtain a frequency domain dictionary matrix and a time domain coefficient matrix, and the cross-correlation function is obtained in combination with the frequency domain weighting function; the time difference from the target noise source to the target UAV sound source collection device is determined according to the cross-correlation function, and the spatial position of the target noise source is determined according to the coordinate relationship; at the same time, according to the cross-correlation function, the amplitude spectrum matrix is ​​filtered in combination with a binary mask matrix based on a preset interference judgment rule to obtain an amplitude spectrum matrix after filtering out noise. The noise source extraction and positioning method and system proposed in the present invention effectively improves positioning accuracy, reduces computational complexity, enhances anti-interference ability, realizes rapid positioning, and has a wide range of applications through technical means such as frequency domain weighting function, matrix decomposition technology, and a binary mask matrix based on a preset interference judgment rule. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0054] Figure 1 A method flow chart of a noise source extraction and positioning method and system provided by an embodiment of the present invention;

[0055] Figure 2 A mixed sound waveform diagram including a noise source extraction and positioning method and system provided by one embodiment of the present invention;

[0056] Figure 3 A binary mask dot matrix diagram including a noise source extraction and location method and system provided by one embodiment of the present invention;

[0057] Figure 4 A transformer sound waveform diagram after noise is filtered out, including a noise source extraction and location method and system provided by one embodiment of the present invention;

[0058] Figure 5 A transformer real sound waveform diagram including a noise source extraction and location method and system provided by one embodiment of the present invention;

[0059] Figure 6An internal structural diagram of a computer device including a noise source extraction and positioning method and system provided in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0061] Example 1

[0062] Reference Figure 1-Figure 6 , which is the first embodiment of the present invention, provides a noise source extraction and positioning method and system, including:

[0063] Before describing the embodiments of the present application in detail, some related concepts are first explained for the sake of clarity.

[0064] Matrix factorization: Matrix factorization is a method of representing a matrix as the product of two or more smaller matrices. In signal processing and machine learning, matrix factorization is often used for tasks such as dimensionality reduction, data compression, and feature extraction. Matrix factorization is used to decompose the amplitude spectrum matrix into a frequency domain dictionary matrix and a time domain coefficient matrix. This decomposition helps to extract meaningful features from the original signal and can better separate the information of different signal sources.

[0065] Frequency Domain Dictionary Matrix: The frequency domain dictionary matrix (Dictionary Matrix in Frequency Domain) is a matrix composed of a set of frequency domain atoms, which represent the basic shapes of different frequency components. The frequency domain dictionary matrix is ​​usually used in signal representation, sparse coding and other scenarios. The frequency domain dictionary matrix is ​​used here to represent the basic forms of different frequency components in the sound signal. By decomposing the amplitude spectrum matrix, a frequency domain dictionary matrix containing basic frequency components can be obtained, which helps to understand the composition structure of the signal.

[0066] Time domain coefficient matrix: The temporal coefficient matrix contains the coefficient information of the signal on the time axis, which describes the activity of each frequency domain atom at different time points. The temporal coefficient matrix here represents the change of different frequency components over time. The temporal coefficient matrix obtained by matrix decomposition can help understand how each frequency component changes over time, so as to further analyze and extract useful signals.

[0067] Binary Mask Matrix: A binary mask matrix is ​​a special matrix whose elements can only be 0 or 1, and is used to select or suppress certain signal components. In signal processing, binary mask matrices are usually used for tasks such as filtering and noise removal. In this method, the binary mask matrix is ​​used to distinguish the target signal from the interference signal. By comparing the time difference obtained under different conditions, it is possible to determine which frequency components belong to the target signal and which belong to the noise, and based on this, a binary mask matrix is ​​constructed to filter out the noise.

[0068] Short-time Fourier transform: Short-time Fourier transform (STFT) is a method to convert time domain signals into frequency domain signals, but it is performed on a short time window of the signal. STFT provides spectral information of the signal at different time points, so it can capture the characteristics of the signal changing over time. Short-time Fourier transform is first used to convert acoustic signals from time domain to frequency domain for subsequent processing. After noise removal, the inverse transform of STFT (Inverse Short-Time Fourier Transform, ISTFT) is also required to convert the filtered frequency domain signal back to the time domain in order to obtain the original signal after denoising.

[0069] There are some problems in the related technology, such as low positioning accuracy, high computational complexity, poor anti-interference ability and slow positioning speed.

[0070] The present application provides a method for effectively solving the above-mentioned problems. Next, multiple embodiments will be combined to explain in detail how to implement the noise source extraction and positioning method.

[0071] Figure 1 A method flow chart of a noise source extraction and positioning method and system is shown, including:

[0072] S101, obtaining a target mixed sound signal collected by the left and right sound source collection devices of the target UAV, converting the target mixed sound signal into a frequency domain signal, and obtaining an amplitude spectrum matrix, the amplitude spectrum matrix including a left amplitude spectrum matrix and a right amplitude spectrum matrix;

[0073] In the embodiment of the present application, two microphones are installed on the rotorcraft drone to collect transformer sound signals, and two groups of mixed sound signals including drone self-noise, environmental noise and transformer sound are collected. The two groups of time domain mixed sound signals are converted into frequency domain signals using short-time Fourier transform, and the amplitude spectrum matrix V of the signal at different time and frequency is obtained by combining its time domain information. l，ft and V r，ft, that is, the left amplitude spectrum matrix and the right amplitude spectrum matrix.

[0074] In the embodiment of the present application, the process of extracting the frequency domain signal from the sound signal collected by the left and right microphones carried by the rotorcraft drone and obtaining the amplitude spectrum matrix is ​​performed according to the following steps:

[0075] 1. Collect time domain signals: First, you need to collect time domain sound signals from the left and right microphones. Assume that the signal collected by the left microphone is s l (t), the signal collected by the right microphone is s r (t).

[0076] 2. Short-time Fourier transform (STFT): Short-time Fourier transform (STFT) is used to convert the time domain signal into a frequency domain signal. STFT is performed by applying a moving window function at different time points. This application will use a Hamming window with a window size of N. Technicians can also use other types of window functions w(n) to window the signal and then perform Fourier transform on the windowed signal.

[0077] 3. Amplitude spectrum matrix: The amplitude spectrum matrix V of the frequency domain signal will be obtained l,ft and V r,ft .

[0078] Exemplary:

[0079] 1. Collect time domain signals, assuming the sampling frequency is f s , the signal duration is T, then the total number of sampling points is N = f s T. The left microphone signal is s l (t), the right microphone signal is s r (t).

[0080] 2. Short-time Fourier transform (STFT), windowing and segmentation: Assume that the signal is windowed using the Hamming window function w(n). The window length is N w , the overlap ratio between windows is p%, then the window step length L=

[0081]

[0082] Calculate the STFT: For the left microphone signal s l (t), STFT can be expressed as:

[0083]

[0084] Where f is the frequency point, t is the time frame, and n is tIndicates the starting sampling point corresponding to the current time frame, j is an imaginary unit.

[0085] Calculate the STFT of the right microphone signal: Similarly, for the right microphone signal s r (t), the STFT expression is:

[0086]

[0087] 3. Amplitude spectrum matrix, the amplitude spectrum is extracted from the complex STFT. For the amplitude spectrum matrix V of the left microphone signal l,ft And the amplitude spectrum matrix V of the right microphone signal r,ft , which can be expressed as:

[0088] V l ,ft(f,t)=|S l (f,t)|V r ,ft(f,t)=|S r (f,t)|

[0089] Among them, |·| represents the modulus of a complex number. In summary, the above steps can be used to extract the frequency domain signal from the sound signals collected by the left and right microphones, and obtain the amplitude spectrum matrix V l,ft and V r,ft .

[0090] It should be noted that the drone can also be other types of aircraft, such as fixed-wing drones, helicopters, etc., or ground or underwater mobile platforms, depending on the requirements of the application scenario. In addition, the collection device can also be other types of sensors, such as sonar, radar, etc., as long as they can collect the target sound signal.

[0091] It should be noted that relevant technical personnel can set more than two sound source collection devices according to their own needs, and also achieve the extraction and positioning of the final noise source through the method of the present application.

[0092] S102, performing matrix decomposition on the amplitude spectrum matrix to obtain a frequency domain dictionary matrix and a time domain coefficient matrix, and combining the frequency domain weighting function to obtain a cross-correlation function;

[0093] In the embodiment of the present application, performing matrix decomposition on the amplitude spectrum matrix includes:

[0094] Decompose the amplitude spectrum matrix into a frequency domain dictionary matrix and a time domain coefficient matrix;

[0095] The frequency domain dictionary matrix is ​​obtained by the following expression:

[0096]

[0097] The time domain coefficient matrix is ​​obtained by the following expression:

[0098]

[0099] Among them, ⊙ represents the multiplication of the corresponding elements of the matrix, α is the sparsity control parameter, V0 represents the amplitude spectrum matrix, that is, the left amplitude spectrum matrix or the right amplitude spectrum matrix, H n Represents the time domain coefficient matrix of the nth magnitude spectrum matrix, W n The frequency domain dictionary matrix representing the nth magnitude spectrum matrix, H n T represents the transpose of the time domain coefficient matrix of the nth magnitude spectrum matrix, W n+1 The frequency domain dictionary matrix representing the n+1th magnitude spectrum matrix, H n+1 The time domain coefficient matrix of the n+1th magnitude spectrum matrix.

[0100] In the present embodiment, V l,ft and V r,ft Decomposed into the left frequency domain dictionary matrix and the right frequency domain dictionary matrix W l，fd , W r，fd and the left time domain coefficient matrix and the right time domain coefficient matrix H l，dt , H r，dt , where d is called the dictionary atom and represents the basis spectrum.

[0101] Exemplary:

[0102]

[0103]

[0104]

[0105]

[0106] In the embodiment of the present application, matrix decomposition is achieved by continuously iterating matrices W and H, and the iteration ends when the number of iterations reaches a set upper limit or the square error is less than a threshold. The overfitting of the matrix decomposition model is reduced by controlling the value of α. When setting the number of dictionary atoms, it is necessary to balance the complexity of the signal and the computing resources. The sparsity control parameter α can select the optimal parameter after evaluating the decomposition effect under different values ​​through a cross-validation method.

[0107] It should be noted that the benefit of decomposing into the above two matrices is that the complex frequency domain signal can be decomposed into a simpler combination of basis functions, which is convenient for subsequent signal processing and analysis. Through the combination of the frequency domain dictionary matrix and the time domain coefficient matrix, the characteristics of the original signal can be effectively represented and data redundancy can be reduced to a certain extent.

[0108] In the embodiment of the present application, obtaining the cross-correlation function in combination with the frequency domain weighted function includes:

[0109] Since the dictionary matrix obtained by matrix decomposition is a non-negative function of frequency, the influence of the spectrum amplitude is eliminated by normalizing the amplitude spectrum of the left and right channels, and each frequency component is normalized to reflect the relative importance of the frequency in the dictionary atom. The signal is filtered using a frequency domain weighting function to highlight the spectral components of the response signal part and suppress the spectrum of the noise part;

[0110]

[0111] Among them, W f Represents the row vector corresponding to frequency f in the dictionary matrix W, V l Represents the left amplitude spectrum matrix, that is, V l,ft , V r Represents the right amplitude spectrum matrix, that is, V r,ft , W represents the frequency domain dictionary matrix of the amplitude spectrum matrix. The frequency domain dictionary matrix includes the left frequency domain dictionary matrix and the right frequency domain dictionary matrix. The division in the formula is the division of the corresponding elements of the matrix.

[0112] In the embodiment of the present application, the cross-correlation function includes:

[0113] According to the left frequency domain dictionary matrix and the right frequency domain dictionary matrix, combined with the frequency domain weighting function, the left frequency domain weighting function and the right frequency domain weighting function of the target mixed sound signal collected by the left sound source collection device are calculated, and the mean of the left frequency domain weighting function and the right frequency domain weighting function are obtained; the cross-correlation function is obtained by weighted summation of all frequency components and combining the amplitude spectra of the left and right channels and their phase relationship.

[0114] In the embodiment of the present application, the left frequency domain weighting function and the right frequency domain weighting function are expressed as ψ l，fd and ψ r，fd ,in:

[0115]

[0116]

[0117] It should be noted that the mean ψ f That is (ψ l,fd +ψ r,fd ) / 2.

[0118] The cross-correlation function is expressed as:

[0119] G τd =Re∑ f ψ f ⊙V l ⊙V r* e j2πfτ

[0120] Where * represents the complex conjugate of the matrix, and the cross-correlation function reflects the function of the sound arrival time difference τ corresponding to each dictionary atom d at each time t, ψ f represents the mean of the left frequency domain weighted function and the right frequency domain weighted function, ⊙ represents the multiplication of the corresponding elements of the matrix, Re represents the real part, e j2 π f τ represents a complex unit.

[0121] It should be noted that the benefit of calculating the cross-correlation function in this way is that it can effectively extract the arrival time difference of the target sound signal at different time points, thereby providing an important basis for subsequent sound source localization. Through the cross-correlation function, the sound signals collected by different sound source collection devices can be time-aligned to further improve the accuracy of sound source localization.

[0122] S103, determining the time difference of the target noise source propagating to the target UAV sound source collection device according to the cross-correlation function, and determining the spatial position of the target noise source according to the coordinate relationship;

[0123] In the embodiment of the present application, determining the time difference of the target noise source propagating to the target UAV sound source collection device according to the cross-correlation function, and determining the spatial position of the target noise source according to the coordinate relationship includes:

[0124] The maximum value of the cross-correlation function in each time period is recorded as the first time difference, where the maximum value is the time difference between the target sound propagating to the left and right sound source collection devices;

[0125] The spatial position of the target noise source is obtained based on the spatial coordinates of the left and right sound source collection devices and the target sound arrival time difference.

[0126] In an optional embodiment, the cross-correlation function G τd The corresponding maximum value in each time period is recorded as the first time difference τ1, where the maximum value is the time difference between the target sound propagating to the left and right sound source collection devices.

[0127]

[0128] Among them: argmax means to find the variable value that makes the function value maximum.

[0129] Furthermore, the spatial position of the target noise source is determined according to the coordinate relationship. Assume that the coordinates of the left and right sound source collection devices are P l (x l ,y l ,z l ) and P r(x r ,y r ,z r ).

[0130] Furthermore, the azimuth angle φ and elevation angle θ of the sound source can be determined: the binaural positioning principle can be used to calculate the azimuth and elevation angles of the target noise source relative to the drone. Given the time difference τ1 and the speed of sound c (assuming it is 343 m / s), the distance difference Δd of the sound source relative to the two microphones can be calculated:

[0131] Δd=c·τ1

[0132] In this embodiment, the binaural positioning formula is used to obtain the azimuth angle φ and the elevation angle θ:

[0133]

[0134]

[0135] Furthermore, the distance d of the sound source is determined: using the distance calculation formula in the binaural positioning principle, the distance d between the sound source and the drone can be obtained:

[0136]

[0137] Furthermore, the source coordinates P are determined. s (x s ,y s ,z s ): The coordinates of the sound source can be determined based on the azimuth angle φ, elevation angle θ and distance d:

[0138] x s =x l +d·cos(φ)·sin(θ)

[0139] y s =y l +d·sin(φ)·sin(θ)

[0140] z s =z l +d·cos(θ)

[0141] It should be noted that the distance to the sound source can also be estimated by measuring the intensity and attenuation of the sound signal. A high-density microphone array can be used to capture the distribution of the sound field and generate a two-dimensional or three-dimensional image of the sound source. The spatial position can also be determined by measuring the frequency change caused by the relative motion between the sound source and the receiver to determine the speed and direction of the sound source.

[0142] S104, filtering the amplitude spectrum matrix according to the cross-correlation function and the binary mask matrix based on the preset interference judgment rule to obtain the amplitude spectrum matrix after the noise is filtered out.

[0143] In the embodiment of the present application, the binary mask matrix based on the preset interference judgment rule includes:

[0144] When the base spectrum is not considered, the frequency domain weighting function is expressed as follows:

[0145]

[0146] Recalculate the cross-correlation function according to the frequency domain weighting function without considering the base spectrum, and the arrival time difference corresponding to the maximum value of the cross-correlation function, which is recorded as the second time difference;

[0147] If the difference between the first time difference and the second time difference is within a preset threshold, the base spectrum is considered to belong to the target signal, otherwise it is related to interference;

[0148] Use the binary mask matrix M dt This association is shown as follows:

[0149]

[0150] Among them, τ dt represents the first time difference, τ t represents the second time difference, ε represents the preset threshold;

[0151] Through the binary mask matrix M dt Get the amplitude spectrum matrix V after filtering out noise ft ', expressed as:

[0152] V ft '=W fd (H dt ⊙M dt )

[0153] Among them, W fd Represents the frequency domain dictionary matrix, including the left frequency domain dictionary matrix W l，fd , right frequency domain dictionary matrix W r，fd ;H dt Represents the time domain coefficient matrix, including the left time domain coefficient matrix H l，dt , right time domain coefficient matrix H r，dt , d is the dictionary atom, representing the base spectrum.

[0154] In the embodiment of the present application, the preset threshold is generally set to a smaller value, such as 0.01 seconds, to ensure high sensitivity when judging whether the base spectrum belongs to the target signal. In this way, the interference signal can be effectively filtered out and the accuracy of sound source positioning can be improved.

[0155] In the embodiment of the present application, the amplitude spectrum matrix after noise filtering is subjected to an inverse short-time Fourier transform, and then the time domain signals are superimposed to obtain the extracted denoised time domain target sound.

[0156] It should be noted that the present application can effectively extract the target sound signal from a complex sound field environment, thereby improving the accuracy and reliability of sound source localization. By combining the cross-correlation function and the binary mask matrix, the interference signal can be effectively suppressed to ensure the purity of the target signal. In addition, by inverse transformation of the short-time Fourier transform and superposition of time domain signals, the denoising effect can be further improved, making the extracted target sound clearer.

[0157] In practical applications, this technology can be widely used in drones, robots, intelligent monitoring and other fields to improve the equipment's ability to identify and locate sound sources. For example, during the flight of a drone, by accurately locating the sound source, it is possible to quickly locate and track specific targets, improving the efficiency and safety of task execution. In an intelligent monitoring system, by accurately identifying and locating the sound source, it is possible to respond to abnormal sounds in a timely manner, improving the sensitivity and accuracy of the monitoring system.

[0158] In addition, this technology can also be combined with other sensor data, such as visual, infrared and other sensor data, to achieve multi-modal data fusion, further improving the accuracy and reliability of sound source positioning. Through data fusion technology, more comprehensive analysis and identification of sound sources in complex environments can be achieved, thereby improving the overall performance of the system.

[0159] In an optional real-time, the sound source signal can also be analyzed and processed in real time through real-time data stream processing technology. Advanced algorithms such as machine learning and deep learning can further improve the accuracy and response speed of sound source positioning. By training the model to identify the characteristics of different sound sources, the system can quickly extract the target sound from the noise and update the sound source location information in real time.

[0160] In drone applications, real-time sound source localization technology can be combined with flight control systems to achieve autonomous obstacle avoidance and target tracking. For example, when a drone is flying in a complex environment, the system can monitor the surrounding sound signals in real time, identify potential obstacles or targets, and automatically adjust the flight path to avoid collision or track the target.

[0161] In the intelligent monitoring system, real-time sound source localization technology can be combined with the video monitoring system to achieve synchronous analysis of sound and image. When the monitoring system detects an abnormal sound, it can immediately start video recording and match the sound source location with the video image, so as to quickly locate the sound source and improve the efficiency of emergency response.

[0162] In addition, real-time sound source localization technology can also be applied to the field of public safety. For example, in large-scale events or public places, by real-time monitoring and analysis of sound source signals, emergencies such as crowd riots and gunshots can be discovered and located in a timely manner, thereby providing strong support for emergency response.

[0163] In practical applications, real-time sound source localization technology can also be combined with other smart devices, such as smart homes and smart cars, to achieve sound source control and interaction. For example, in a smart home system, by locating the sound source in real time, voice control of a specific area can be achieved, improving the convenience and intelligence of the user experience.

[0164] In short, real-time sound source localization technology has broad application prospects in many fields. By continuously optimizing algorithms and improving data processing capabilities, the performance of sound source localization technology can be further improved, providing more accurate, fast and reliable sound source localization solutions for various application scenarios.

[0165] In summary, the present invention proposes a method for extracting and locating a noise source, obtaining a target mixed sound signal collected by the left and right sound source collection devices of the target UAV, converting the target mixed sound signal into a frequency domain signal, and obtaining an amplitude spectrum matrix, the amplitude spectrum matrix including a left amplitude spectrum matrix and a right amplitude spectrum matrix; performing matrix decomposition on the amplitude spectrum matrix to obtain a frequency domain dictionary matrix and a time domain coefficient matrix, and combining the frequency domain weighting function to obtain a cross-correlation function; determining the time difference of the target noise source propagating to the target UAV sound source collection device according to the cross-correlation function, and determining the spatial position of the target noise source according to the coordinate relationship; at the same time, according to the cross-correlation function, combining a binary mask matrix based on a preset interference judgment rule, filtering the amplitude spectrum matrix to obtain an amplitude spectrum matrix after filtering out noise. The noise source extraction and positioning method and system proposed in the present invention effectively improves positioning accuracy, reduces computational complexity, enhances anti-interference ability, realizes rapid positioning, and has a wide range of applications through technical means such as frequency domain weighting function, matrix decomposition technology, and a binary mask matrix based on a preset interference judgment rule.

[0166] This embodiment also provides a noise source extraction and positioning system, including:

[0167] The amplitude spectrum acquisition module is used to acquire the target mixed sound signal collected by the left and right sound source acquisition devices of the target UAV, convert the target mixed sound signal into a frequency domain signal, and obtain an amplitude spectrum matrix, which includes a left amplitude spectrum matrix and a right amplitude spectrum matrix;

[0168] The cross-correlation function calculation module is used to perform matrix decomposition on the amplitude spectrum matrix to obtain the frequency domain dictionary matrix and the time domain coefficient matrix, and obtain the cross-correlation function in combination with the frequency domain weighting function;

[0169] A positioning module is used to determine the time difference from the target noise source to the target UAV sound source collection device according to the cross-correlation function, and to determine the spatial position of the target noise source according to the coordinate relationship;

[0170] The extraction module is used to filter the amplitude spectrum matrix according to the cross-correlation function and the binary mask matrix based on the preset interference judgment rule to obtain the amplitude spectrum matrix after the noise is filtered out.

[0171] The above-mentioned unit modules may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above-mentioned modules.

[0172] This embodiment also provides a computer device, which may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 6 As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for extracting and locating a noise source is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0173] This embodiment further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0174] Obtain the target mixed sound signal collected by the left and right sound source collection devices of the target UAV, convert the target mixed sound signal into a frequency domain signal, and obtain an amplitude spectrum matrix, which includes a left amplitude spectrum matrix and a right amplitude spectrum matrix;

[0175] Perform matrix decomposition on the amplitude spectrum matrix to obtain the frequency domain dictionary matrix and the time domain coefficient matrix, and combine the frequency domain weighting function to obtain the cross-correlation function;

[0176] Determine the time difference from the target noise source to the target UAV sound source collection device based on the cross-correlation function, and determine the spatial position of the target noise source based on the coordinate relationship;

[0177] At the same time, according to the cross-correlation function and combined with the binary mask matrix based on the preset interference judgment rule, the amplitude spectrum matrix is ​​filtered to obtain the amplitude spectrum matrix after the noise is filtered out.

[0178] Example 2

[0179] Reference Figure 2-Figure 5 , is an embodiment of the present invention, and provides a noise source extraction and positioning method and system. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through experiments, and the steps are as follows:

[0180] Step 1: Two microphones are installed on the rotorcraft drone to collect transformer sound signals. Two sets of mixed sound signals including drone self-noise, environmental noise and transformer sound are collected, such as Figure 2 As shown. Use short-time Fourier transform to convert the time domain mixed signal collected by the left and right microphones into a frequency domain signal, and combine its time domain information to obtain the amplitude spectrum matrix V of the signal at different times and frequencies. l,ft and V r,ft .

[0181] Step 2: Substitute the signal matrix V obtained in step 1 l,ft and V r,ft Decompose into frequency domain dictionary matrix W through matrix decomposition l，fd , W r，fd and the time domain coefficient matrix H l，dt , H r，dt Taking into account the overall perceptual quality, the perceptual quality of the target sound and the perceptual degree of the interference signal, 120 dictionary atoms were used, the sparsity control parameter was set to 0.1, and 230 iterations were performed.

[0182] Step 3: Perform frequency domain normalization on the frequency domain dictionary matrix obtained in step 2 to reflect the relative importance of each frequency component in the dictionary atom. Use the frequency domain weighting function to filter the signal, highlight the spectral components of the response signal part, and suppress the noise components.

[0183] Step 4: Use the dictionary matrix W obtained in step 2 l，fd , W r，fd , calculate the frequency weighting function ψ of the left and right microphone signals l，fd and ψ r，fd , taking the average value to get ψ f The cross-correlation function is calculated by weighted summing of all frequency components and combining the amplitude spectra of the left and right channels and their phase relationship.

[0184] Step 5: Cross-correlation function G τd The maximum value corresponding to each time t is denoted as τ dt , where the maximum value is the time difference between the target sound propagating to the left and right microphones. Combining the spatial coordinates of the microphones and the arrival time difference of the target sound, the spatial position of the target sound source can be determined.

[0185] Step 6: When the dictionary atoms are not considered, the frequency domain weighting function ψ can be obtained f0 , substituting into the cross-correlation function to obtain G τd , the corresponding arrival time difference when taking the maximum value is recorded as τ t . When the dictionary atom corresponds to τ dt With the estimated τ t If the difference is within a certain range, the atom is considered to belong to the target signal, otherwise it is related to interference. dt To show this relationship, Figure 3 A binary mask dot pattern is shown.

[0186] Step 7: Obtain the amplitude spectrum matrix V after filtering out noise through the mask matrix ft '. Perform the inverse short-time Fourier transform on the updated amplitude spectrum matrix (the window function is the same as when adding a window), and then superimpose the time domain signals (take the average of the overlapping parts of the two frames) to obtain the denoised time domain target sound, such as Figure 4 shown. Figure 5 This is the real sound waveform of the transformer. After comparison, it is found that the use of this algorithm can obtain results close to the real value.

[0187] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0188] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0189] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0190] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0191] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0192] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0193] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for extracting and locating noise sources, characterized in that: include: Acquire the target mixed sound signal collected by the left and right sound source collection devices of the target UAV, convert the target mixed sound signal into a frequency domain signal, and obtain an amplitude spectrum matrix, wherein the amplitude spectrum matrix includes a left amplitude spectrum matrix and a right amplitude spectrum matrix; Performing matrix decomposition on the amplitude spectrum matrix to obtain a frequency domain dictionary matrix and a time domain coefficient matrix, and obtaining a cross-correlation function in combination with a frequency domain weighting function; Determine the time difference of the target noise source propagating to the target UAV sound source collection device according to the cross-correlation function, and determine the spatial position of the target noise source according to the coordinate relationship; At the same time, according to the cross-correlation function, combined with a binary mask matrix based on a preset interference judgment rule, the amplitude spectrum matrix is ​​filtered to obtain the amplitude spectrum matrix after the noise is filtered out.

2. The noise source extraction and positioning method according to claim 1, characterized in that: Also includes: The inverse short-time Fourier transform is performed on the amplitude spectrum matrix after noise filtering, and then the time domain signals are superimposed to obtain the extracted denoised time domain target sound.

3. The noise source extraction and positioning method according to claim 2, characterized in that: The matrix decomposition of the amplitude spectrum matrix comprises: Decomposing the amplitude spectrum matrix into a frequency domain dictionary matrix and a time domain coefficient matrix; The frequency domain dictionary matrix is ​​obtained by the following expression: The time domain coefficient matrix is ​​obtained by the following expression: Among them, ⊙ represents the multiplication of the corresponding elements of the matrix, α is the sparsity control parameter, V0 represents the amplitude spectrum matrix, that is, the left amplitude spectrum matrix or the right amplitude spectrum matrix, H n Represents the time domain coefficient matrix of the nth magnitude spectrum matrix, W n The frequency domain dictionary matrix representing the nth magnitude spectrum matrix, H n T represents the transpose of the time domain coefficient matrix of the nth magnitude spectrum matrix, W n+1 The frequency domain dictionary matrix representing the n+1th magnitude spectrum matrix, H n+1 The time domain coefficient matrix of the n+1th magnitude spectrum matrix.

4. The noise source extraction and positioning method according to claim 3, characterized in that: The step of obtaining the cross-correlation function by combining the frequency domain weighted function comprises: The signal is filtered using a frequency domain weighting function to highlight the spectrum components of the response signal part and suppress the spectrum of the noise part; Among them, W f Represents the row vector corresponding to frequency f in the dictionary matrix W, V l represents the left amplitude spectrum matrix, V r represents the right amplitude spectrum matrix, W represents the frequency domain dictionary matrix of the amplitude spectrum matrix, and the frequency domain dictionary matrix includes the left frequency domain dictionary matrix and the right frequency domain dictionary matrix.

5. The noise source extraction and positioning method according to claim 4, characterized in that: The cross-correlation function comprises: According to the left frequency domain dictionary matrix and the right frequency domain dictionary matrix, in combination with the frequency domain weighting function, the left frequency domain weighting function and the right frequency domain weighting function of the target mixed sound signal collected by the left sound source collection device are calculated, and the mean of the left frequency domain weighting function and the right frequency domain weighting function is obtained; The cross-correlation function is expressed as: Where * represents the complex conjugate of the matrix, and the cross-correlation function reflects the function of the sound arrival time difference τ corresponding to each dictionary atom d at each time t, ψ f represents the mean of the left frequency domain weighted function and the right frequency domain weighted function, ⊙ represents the multiplication of the corresponding elements of the matrix, Re represents the real part, e j2πfτ Indicates plural units.

6. The noise source extraction and positioning method according to claim 5, characterized in that: The step of determining the time difference of the target noise source propagating to the target UAV sound source collection device according to the cross-correlation function, and determining the spatial position of the target noise source according to the coordinate relationship includes: The maximum value of the cross-correlation function in each time period is recorded as the first time difference, where the maximum value is the time difference between the target sound propagating to the left and right sound source collection devices; The spatial position of the target noise source is obtained based on the spatial coordinates of the left and right sound source collection devices and the target sound arrival time difference.

7. The noise source extraction and positioning method according to claim 6, characterized in that: The binary mask matrix based on the preset interference judgment rule includes: When the base spectrum is not considered, the frequency domain weighting function is expressed as follows: Recalculate the cross-correlation function according to the frequency domain weighting function without considering the base spectrum, and the arrival time difference corresponding to the maximum value of the cross-correlation function, which is recorded as the second time difference; If the difference between the first time difference and the second time difference is within a preset threshold, the base spectrum is considered to belong to the target signal, otherwise it is related to interference; Use the binary mask matrix M dt This association is shown as follows: Among them, τ dt represents the first time difference, τ t represents the second time difference, ε represents the preset threshold; Through the binary mask matrix M dt Get the amplitude spectrum matrix V after filtering out noise ft ', expressed as: V ft '=W fd (H dt ⊙M dt ) Among them, W fd Represents the frequency domain dictionary matrix, including the left frequency domain dictionary matrix W l，fd , right frequency domain dictionary matrix W r，fd ;H dt Represents the time domain coefficient matrix, including the left time domain coefficient matrix H l，dt , right time domain coefficient matrix H r，dt , d is the dictionary atom, representing the base spectrum.

8. A noise source extraction and positioning system, characterized in that: include: An amplitude spectrum acquisition module is used to acquire the target mixed sound signal collected by the left and right sound source acquisition devices of the target UAV, convert the target mixed sound signal into a frequency domain signal, and obtain an amplitude spectrum matrix, wherein the amplitude spectrum matrix includes a left amplitude spectrum matrix and a right amplitude spectrum matrix; A cross-correlation function calculation module, used to perform matrix decomposition on the amplitude spectrum matrix to obtain a frequency domain dictionary matrix and a time domain coefficient matrix, and obtain a cross-correlation function in combination with a frequency domain weighting function; A positioning module, used to determine the time difference of the target noise source propagating to the target UAV sound source collection device according to the cross-correlation function, and determine the spatial position of the target noise source according to the coordinate relationship; The extraction module is used to filter the amplitude spectrum matrix according to the cross-correlation function and the binary mask matrix based on the preset interference judgment rule to obtain the amplitude spectrum matrix after the noise is filtered out.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images