A multi-path underwater acoustic signal separation method based on template matching filtering

Abstract

The present invention discloses a method for separating multi-path hydroacoustic signals based on template matching filtering, comprising the following steps: (1) performing template matching filtering processing on hydroacoustic signals received from multiple paths to obtain a matching filtering result of a mixed signal; using a single sound wave signal as a template, performing template matching filtering calculation on the signal itself to obtain a single signal matching filtering result; (2) using the single signal matching filtering result as a template, constructing a template matrix, and performing signal processing; (3) obtaining a result of separating multi-path hydroacoustic signals through a linear fitting algorithm with constraints according to the result of the template matrix processing; (4) using the shortest path matching algorithm to connect the arrival time points of hydroacoustic signals belonging to the same sound line in continuous signal data to form a sound wave propagation time line. The present invention is used in the field of acoustic tomography flow measurement, and can accurately identify and separate different hydroacoustic signals, thereby analyzing the cross-section of the river channel and improving the measurement accuracy.

Description

Technical Field

[0001] The invention relates to the field of intelligent analysis of river flow measurement data, and specifically refers to a method for separating multi-path acoustic wave signals in water by performing calculation processing on received mixed acoustic wave signals in the field of acoustic tomography flow measurement. Background Art

[0002] River flow measurement is one of the most basic and important measurement tasks in hydrological and water conservancy projects. The flow measurement data of river basins is the main basis for water conservancy and hydrological surveys, water conservancy project design, flood control command decisions, and water resources scheduling. The river acoustic tomography flow measurement technology calculates the river flow rate based on the characteristic that the propagation speed of sound waves in the downstream is faster than that in the upstream. Therefore, accurately calculating the propagation speed of sound waves in water is the key to acoustic tomography flow measurement. Since sound waves will reach the receiving device through different propagation paths (sound lines) in the water body, the received signal is the superposition of signals from multiple paths. At the same time, the received signal will also be interfered by water body noise and equipment noise, such as the undulation of the riverbed and the influence of different media such as silt, sand, gravel, and aquatic plants that make up the riverbed. This will make it difficult to identify and distinguish the signals of different sound lines, causing large errors in the calculation of the propagation time of sound waves, and affecting the subsequent calculation of the flow rate and flow of the river.

[0003] At present, the unmodulated underwater acoustic signal is usually identified by the signal-to-noise ratio, while the coded modulated signal is calculated by the filtering algorithm to obtain the signal matching value, and then the signal position is found by finding the maximum value. However, this traditional method has some problems. First, the signal matching value is related to the signal strength, and the signal corresponding to the maximum value is not necessarily the best signal for detection; second, when multiple signals and noise are superimposed, pseudo peak points may appear, causing misidentification and ultimately leading to errors in the calculation results. Existing patent technology: A mixed underwater acoustic signal separation method and system combining auditory scene analysis and deep learning, patent application number CN202410161091.0, the method includes: performing auditory segmentation on the acquired mixed underwater acoustic signal through a pre-trained auditory segmentation model, and decomposing the mixed underwater acoustic signal into a number of auditory segments; wherein each of the auditory segments is used to characterize the local description of the corresponding acoustic event in the auditory scene; performing auditory reorganization on the several auditory segments through a pre-trained deep clustering network, recombining the auditory segments from the same sound source together, and obtaining the underwater acoustic signal corresponding to each of the sound sources, which can realize the separation and reconstruction of mixed underwater acoustic signals in complex underwater environments. This method decomposes the underwater acoustic signal into a number of auditory segments. The algorithm process for processing multi-path mixed underwater acoustic signals is complex, and the requirements for cutting and restoring the time nodes of multiple auditory segments are very demanding, and it is difficult to accurately separate the signal. Summary of the invention

[0004] In order to solve the problem that the hydroacoustic sound wave signals cannot be correctly identified and separated in the river acoustic tomography flow measurement method, the present invention proposes a multi-path hydroacoustic signal separation method based on template matching filtering, which can accurately identify and separate different hydroacoustic sound wave signals, thereby analyzing the cross-section of the river and realizing the measurement of the entire cross-section of the river.

[0005] The present invention provides a multi-path underwater acoustic signal separation method based on template matching filtering, comprising the steps of:

[0006] (1) Performing template matching filtering on the received multi-path underwater acoustic signal to obtain the matched filtering result of the mixed signal; using a single acoustic wave signal as a template, performing template matching filtering calculation on the signal itself to obtain the matched filtering result of the single signal;

[0007] (2) using the single signal matched filtering result as a template, constructing a template matrix, and performing signal processing;

[0008] (3) according to the result of the template matrix processing, a linear fitting algorithm with constraints is used to obtain the result of multi-path underwater acoustic signal separation;

[0009] (4) Using the shortest path matching algorithm, the minimum weight matching algorithm, or the Hungarian algorithm, the arrival time points of the underwater acoustic signals belonging to the same sound line in the continuous signal data are connected to form a sound wave propagation time line.

[0010] Furthermore, in step (1), a single sound wave signal is used as a template to perform template matching filtering calculation on the signal itself to obtain a single signal matched filtering result. By utilizing the consistency between the length of the matched filtering result and the length of the matched data, the curve change value of the area near the maximum value point is intercepted for subsequent processing.

[0011] Further, in step (2), the single signal matched filtering result T is used as a template, and the template matrix is ​​constructed as follows: 1) Assuming that the length of the mixed signal data to be separated is N, an all-zero matrix M of size (N, N) is constructed; 2) The j-th column value of M is replaced by:

[0012] M (:,j) =[0,...,t i-n ,t i-n+1 ,...,t i ,...t i+n-1 ,t i+n ,...,0], where t i At the j-th position, if in<0 or i+n>=N, corresponding truncation processing is performed to finally obtain the template matrix.

[0013] Furthermore, in step (3), assuming that the matched filtering result of the received mixed signal is y, the template matrix is ​​M, and the result of signal separation is A, there is a relationship: y = MA, where y and M are known. Through a linear fitting algorithm with constraints, a multivariate linear fitting is performed on the relationship, and A is obtained, which is the result of the desired underwater acoustic signal separation.

[0014] Furthermore, the result of solving A is obtained by a linear fitting algorithm with constraints, where the value of A is non-negative and only a part of the time points are positive. Each non-zero component of A represents a signal at the corresponding time point, which is consistent with the actual situation of multi-path underwater acoustic signal mixing.

[0015] Furthermore, the signal separation results are processed and the arrival time points of the sound wave signals belonging to the same path are connected to form a sound wave propagation timeline. Each timeline represents the change of a sound line on the river section.

[0016] Compared with the prior art, the technical advantages brought by the method of the present invention are as follows:

[0017] (1) By comparison, the existing hydroacoustic signal filtering algorithm, because the mixed sound wave signal contains water noise and equipment noise, if the traditional filter matrix operation is used to directly solve the separated hydroacoustic signal, serious overfitting will occur, resulting in distortion of the separated signal. Different from the traditional sound signal filtering algorithm, the present invention utilizes the interchangeable characteristics of signal superposition and matched filtering processing to convert the mixed signal separation process into a matched filtering result separation process. First, a template matrix is ​​constructed, and a signal solution formula is introduced. Then, the position of each hydroacoustic signal is calculated through an original linear fitting method with constraints to achieve signal separation, which greatly improves the accuracy. It is 5-7% higher than the traditional signal separation algorithm, with a low distortion rate, and improves the measurement accuracy of river velocity and flow.

[0018] (2) For wide river sections, the present invention also uniquely uses the shortest path matching algorithm, or the minimum weight matching algorithm, or the Hungarian algorithm to connect the arrival time points of the hydroacoustic signals belonging to the same sound line in the continuous signal data to form a sound wave propagation timeline. An intuitive graph displays the changes in multiple sound lines on the section, thereby being able to analyze the propagation of sound lines in the water body and the river section. The method of the present invention can separate multiple sound lines, obtain more water body information, and realize the measurement of the entire river section. The signal processing integration and intelligence level are high, reflecting the latest scientific and technological hydroacoustic signal separation technology.

[0019] (3) The method of the present invention has wider applicability and overcomes the influence of interference noise caused by the fluctuation of the riverbed and the different media such as silt, sand, gravel, and aquatic plants that constitute the riverbed. It has good filtering fidelity, short calculation time, and high stability when used in measurement systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Schematic diagram of the interchangeability of signal superposition and matched filtering processing.

[0021] Figure 2 This is a schematic diagram of the single signal matched filtering results.

[0022] Figure 3 Schematic diagram of the matched filtering results of mixed signals.

[0023] Figure 4 It is a schematic diagram of using existing traditional matrix operations to directly solve the separated signals.

[0024] Figure 5 It is a schematic diagram of solving the separation signal using a linear fitting algorithm with restriction conditions in the method of the present invention.

[0025] Figure 6 It is a schematic diagram of signal reproduction based on the signal separation result.

[0026] Figure 7 It is a schematic diagram of the continuous data calculation results of the method of the present invention.

[0027] Figure 8 The data calculation result of the method of the present invention is Figure 7 A partial enlarged view of .

[0028] Fig. 9 It is a diagram of the sound ray separation result of the method of the present invention. DETAILED DESCRIPTION

[0029] The technical solution in the embodiment of the invention will be clearly and completely described below in conjunction with the drawings in this embodiment.

[0030] The invention discloses a multi-path underwater acoustic signal separation method based on template matching filtering, which uniquely utilizes the interchangeable characteristics of signal superposition and matched filtering processing, and converts the signal separation process into a matched filtering result separation process for processing.

[0031] See also Figure 1 The interchangeability of signal superposition and matched filtering processing is specifically embodied as follows: multiple sound wave signals are first superimposed to obtain a mixed sound wave signal, and then processed by a template matching filtering algorithm to obtain a matched filtering result R1 of the mixed signal; multiple sound wave signals are first processed by a template matching filtering algorithm to obtain multiple matched filtering results, and then superimposed to obtain a matched filtering result R2 of the mixed signal. In theory, R1 and R2 are completely equal. This property can be used to convert the problem of underwater acoustic signal separation into the problem of matching filtering result separation, and then, based on the characteristics of the underwater acoustic signal, a linear fitting method with constraints is used to calculate the position of each signal to achieve accurate signal separation. The implementation process of the method of the present invention is described in detail below.

[0032] The present invention provides a multi-path underwater acoustic signal separation method based on template matching filtering, comprising:

[0033] Step (1): according to the type of underwater acoustic signal sent, generally an M code signal is sent, and template matching filtering is performed on the received multi-path underwater acoustic signal to obtain a matched filtering result of the mixed signal; using a single sound wave signal as a template, template matching filtering calculation is performed on the signal itself to obtain a single signal matched filtering result.

[0034] The single signal matched filtering result is as follows Figure 2 , where the maximum point is exactly where the first point of the signal is. Note that the length of the matched filtering result is consistent with the length of the matched data. When analyzing the signal, only the curve changes in the area near the maximum point are concerned (for example Figure 2 The horizontal axis of the time axis is 975-1025 interval), so when using the matched filtering result, it is sufficient to ensure that its length can contain this interval. Assume that the matched filtering result is T = [t i-n ,t i-n+1 ,...,t i ,...,t i+n-1 ,t i+n ], where t i It is the maximum value point, and n can be greater than 24.

[0035] Step (2), using the single signal matched filtering result as a template, constructing a template matrix, and performing signal processing. Using the above matched filtering result T as a template, the method for constructing the template matrix is ​​as follows: 1) Assuming that the length of the mixed signal data to be separated is N, construct an all-zero matrix M of size (N, N); all zero means no signal, which is used as a background here. On this basis, templates are added to each column to form the final template matrix; 2) The j-th column value of M is replaced by:

[0036] M (:,j) =[0,...,t i-n ,t i-n+1 ,...,t i ,...t i+n-1 ,t i+n ,...,0], where ti is located at the jth position. If in<0 or i+n>=N, corresponding truncation is performed.

[0037] Finally, the template matrix is ​​obtained:

[0038]

[0039] Assume that the matched filtering result of the received mixed signal is y, the template matrix is ​​M, and the result of signal separation is A. A is a non-negative vector of length N. The size of the i-th component value represents the scaled size of the signal at the current position. If the value is equal to 0, it means that there is no signal at this position.

[0040] Step (3), based on the result of the template matrix processing, a linear fitting algorithm with constraints is used to obtain the result of multi-path underwater acoustic signal separation.

[0041] The result of matched filtering is y, the template matrix is ​​M, and the result of signal separation is A. The relationship between the above three is: y = MA. Given y and M, solve A. Since the mixed sound wave signal contains water noise and equipment noise, if the existing traditional matrix operation direct solution method is used to calculate A, serious overfitting will occur, resulting in multi-noise signal interference.

[0042] Therefore, the present invention uses a unique linear fitting algorithm with constraints to solve A. Since the value of A is non-negative and the number of target acoustic wave signals contained in the mixed signal is limited, the value of A should have as many zero components as possible. Each non-zero component of A represents that there is a signal at that time point, and 0 means that there is no signal at that time point. Since the number of superimposed signals is actually limited, A will not have too many non-zero components. Therefore, the present invention uses a multivariate linear fitting method with non-negative value constraints and L1 regularization constraints to solve A, which will be more in line with actual signal conditions.

[0043] The loss function in the restricted linear fitting of the present invention is: L = (MA-y) 2 +λ||A||1(x≥0), where λ is a hyperparameter used to adjust the weight of the L1 regularization term in the loss function, which is related to the size of the signal data. The goal of fitting is to find an A value that minimizes L. It is an iterative process. The A that minimizes L is the desired signal separation result.

[0044] A signal separation embodiment is given below. Figure 3 is the matched filtering result y of the actually received mixed signal. Figure 4 This is the result of directly solving A using existing traditional matrix operations. It can be seen that the values ​​of A are both positive and negative, and the separated signals fill the time axis, which is completely inconsistent with the actual separation of mixed sound wave signals. This is an obvious overfitting phenomenon, resulting in signal distortion.

[0045] Figure 5 This is the result of solving A using a linear fitting method with constraints in the present invention. It can be seen that the value of A is non-negative, and only a portion of time points are positive, which is consistent with the actual situation of sound wave signal mixing. Figure 6Based on the signal separation result A, the single signal matched filtering result is re-superimposed into the mixed signal matched filtering result, and then compared with the original result y, it can be seen that the reproduced signal is highly similar to the original signal. The calculated error level is consistent in the noise interval and the signal interval, which can be considered to be caused by noise. Figure 5 and Figure 6 It can be seen that the method of the present invention has high accuracy and low distortion in separating multi-path underwater acoustic signals.

[0046] The method of the present invention calculates multiple sets of continuous data and obtains the corresponding signal separation results as follows: Figure 7 , Figure 7 The enlarged picture of the calculation results in the middle part is as follows Figure 8 As shown, the brighter the color, the greater the value. In the figure, the horizontal axis is the time axis, and the vertical axis is the arrival time of the sound wave. After calculation, the time point corresponding to the first component of A higher than 2000 obtained by solution is basically consistent with the actual arrival time point of the sound wave, and other sound wave signals have a high degree of continuity and uniformity of change trend over time, proving that the sound wave signal calculated by the algorithm of the present invention is true and reliable, with high accuracy. The above-mentioned 2000 component is a noise threshold. It can be considered that the earliest measured sound wave intensity is relatively high, and the value below 2000 is likely to be noise. The definition of 2000 value is based on a large number of test results and empirical statistics. It can be seen from the distribution of sound wave signals that the water body changes are relatively stable and continuous as a whole, and there are occasional jumps in the middle, such as a ship passing through the water surface, which is within the allowable range of change.

[0047] Step (4), using the shortest path matching algorithm, or the minimum weight matching algorithm, or the Hungarian algorithm, connects the arrival time points of the underwater acoustic signals belonging to the same sound line in the continuous signal data to form a sound wave propagation time line.

[0048] As an implementation method, the signal separation result is calculated by the shortest path matching algorithm, such as Fig. 9 As shown in the figure, the arrival time points of the sound wave signals belonging to the same sound line are connected to form a sound wave propagation time line. Each line in the figure represents the change of a sound line on the section, completing the separation of multi-path underwater acoustic signals.

[0049] The method of the present invention constructs a template matrix, introduces a signal solution formula, and then calculates the position of each hydroacoustic signal through an original linear fitting method with constraints to achieve signal separation. Through comparative test analysis, the accuracy is greatly improved, the distortion rate is small, and the measurement accuracy of river velocity and flow is improved. The scheme of the present invention can intuitively display the changes of a sound line on the section, so as to analyze the propagation of the sound line in the water body and the cross-section of the river, and realize the measurement of the entire cross-section of the river. The signal processing integration and intelligent level are high.

Claims

1. A multi-path underwater acoustic signal separation method based on template matching filtering, characterized in that: Includes steps: (1) Performing template matching filtering on the received multi-path underwater acoustic signal to obtain a matched filtering result of the mixed signal; Using a single sound wave signal as a template, perform template matching filtering calculation on the signal itself to obtain a single signal matched filtering result; (2) using the single signal matched filtering result as a template, constructing a template matrix, and performing signal processing; (3) According to the result of the template matrix processing, the result of multi-path underwater acoustic signal separation is obtained by a linear fitting algorithm with constraints; assuming that the matched filtering result of the received mixed signal is y, the template matrix is ​​M, and the result of signal separation is A, there is a relationship: y = MA, where y and M are known, and a linear fitting algorithm with constraints is used to perform multivariate linear fitting on the relationship, and A is obtained as the result of the underwater acoustic signal separation; the loss function in the constrained linear fitting is: L = (MA-y) 2 +λ‖A‖1(x≥0), where λ is a hyperparameter; (4) Using the shortest path matching algorithm, the minimum weight matching algorithm, or the Hungarian algorithm, the arrival time points of the underwater acoustic signals belonging to the same sound line in the continuous signal data are connected to form a sound wave propagation time line.

2. The multi-path underwater acoustic signal separation method based on template matching filtering according to claim 1 is characterized in that: In step (1), Using a single sound wave signal as a template, a template matching filter calculation is performed on the signal itself to obtain a single signal matched filtering result. Using the consistency between the length of the single signal matched filtering result and the length of the matched data, the curve change value of the area near the maximum value point is intercepted for subsequent processing.

3. The multi-path underwater acoustic signal separation method based on template matching filtering according to claim 1 or 2 is characterized in that: In step (2), the single signal matched filtering result T is used as a template, and the template matrix is ​​constructed as follows: 1) Assuming that the length of the mixed signal data to be separated is N, an all-zero matrix M of size (N, N) is constructed; 2) The j-th column value of M is replaced by: M (:,j) =[0,...,t i-n ,t i-n+1 ,...,t i ,...t i+n-1 ,t i+n ,...,0], where t i At the j-th position, if in<0 or i+n>=N, corresponding truncation processing is performed to finally obtain the template matrix.

4. The multi-path underwater acoustic signal separation method based on template matching filtering according to claim 1 is characterized in that: The result of solving A is obtained by a linear fitting algorithm with constraints, where the value of A is non-negative and only a part of the time points are positive. Each non-zero component of A represents a signal at the corresponding time point, which is consistent with the actual situation of multi-path underwater acoustic signal mixing.

5. The multi-path underwater acoustic signal separation method based on template matching filtering according to claim 1 is characterized in that: The signal separation results are processed by connecting the arrival time points of the sound wave signals belonging to the same path to form a sound wave propagation timeline. Each timeline represents the change of a sound line on the river section.

Images