A nonlinear characteristic detection method for underwater noise

Abstract

The present invention discloses a nonlinear characteristic detection method for underwater noise, comprising: S1, collecting underwater noise signals and performing permutation entropy extraction; S2, performing multi-scale feature classification on the permutation entropy, obtaining the differentiated characteristics of underwater noise, and realizing nonlinear characteristic detection of underwater noise. Aiming at the characteristics of cavitation noise and ocean environmental noise, the present invention utilizes multi-scale phase permutation entropy and improved multi-scale permutation entropy to extract complexity characteristics, realizes the characteristic distinction between cavitation noise and ocean background noise, and improves the detection capability of passive sonar systems.

Description

Technical Field

[0001] The invention belongs to the technical field of noise characteristic analysis, and in particular relates to a nonlinear characteristic detection method for underwater noise. Background Art

[0002] Underwater noise includes ocean background noise, target (ship, torpedo) radiation noise and target self-noise. Among them, ocean background noise and target noise are one of the main interference backgrounds of sonar system, and target radiation noise is the sound source of passive sonar system. Passive sonar system receives this noise to detect the target. Propeller cavitation noise is one of the main noise components of target self-noise and target radiation noise. In-depth analysis of the characteristics of cavitation noise and ocean background noise is of great significance to improving the concealment and safety of the system and improving the detection and identification capabilities of passive systems.

[0003] At present, the characteristic analysis methods for cavitation noise and ocean background noise mainly include line spectrum analysis, high-order spectrum analysis, and time-frequency analysis. However, the underwater ocean environment is complex and changeable. Affected by the random, time-varying, space-varying, and non-uniform ocean transmission medium, the underwater acoustic target signal has strong distortion and nonlinear fluctuations, showing the "three non-stationary, non-Gaussian, and nonlinear" characteristics. Therefore, the traditional signal characteristic analysis method has obvious limitations. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the nonlinear characteristic detection method of underwater noise provided by the present invention solves the problem that the existing underwater acoustic target signal is affected by the underwater marine environment and it is difficult to accurately analyze its characteristic differences.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is: a nonlinear characteristic detection method of underwater noise, comprising the following steps:

[0006] S1, collecting underwater noise signals and extracting information entropy;

[0007] S2, performing multi-scale feature classification on the permutation entropy to obtain differentiated features of underwater noise and realize nonlinear feature detection of underwater noise;

[0008] The underwater noise signal is an ocean background noise signal and a propeller cavitation noise signal of the same voyage, and the information entropy is a phase arrangement entropy or an improved arrangement entropy.

[0009] Furthermore, in step S1, when the information entropy is phase arrangement entropy, the method for extracting the phase arrangement entropy is:

[0010] A1. Determine the instantaneous phase sequence of underwater noise signal;

[0011] A2, coarse-graining the instantaneous phase sequence to obtain an instantaneous phase subsequence;

[0012] A3, reconstructing the instantaneous phase subsequence in phase space to obtain several row vectors;

[0013] A4. Arrange the elements in each row vector in ascending order according to the dimension of the reconstructed phase space to obtain several order patterns;

[0014] A5. Determine the probability of each order pattern appearing in the phase space;

[0015] A6. Determine the phase arrangement entropy based on the probability of occurrence of each order pattern.

[0016] Furthermore, in step A2, the formula for coarse-graining the instantaneous phase sequence is:

[0017]

[0018] In the formula, is the instantaneous phase subsequence y under scale s (s) The elements in , s is the scale factor, x i is the instantaneous phase of the signal, x i ∈X, X is the instantaneous phase sequence, X={x1,x2,x3,...,x N}, N is the total number of instantaneous phases in X, subscript i is the instantaneous phase sequence number, and subscript j is the instantaneous phase subsequence y formed after the coarse-grained processing of sequence X (s) The ordinal number of the element in;

[0019] In step A3, the row vector Y i s for:

[0020]

[0021] In the formula, is the row vector Y i s The elements in , τ is the time delay, and m is the dimension of the reconstructed phase space;

[0022] In step A4, the order pattern obtained by arranging the elements in the row vector in ascending order is:

[0023]

[0024] Where k1, k2, ..., k m is the original position of each element, π i =[k1,k2,...,k m ],π i is a symbol vector;

[0025] In step A5, the probability p of each order pattern appearing l for:

[0026] P l =h l / N-m+1

[0027] In the formula, h l is the frequency of occurrence of each order pattern, subscript l = 1, 2, ..., m!;

[0028] In step A6, the expression of phase arrangement entropy PE(m) is:

[0029]

[0030] Furthermore, the step S2 is specifically as follows:

[0031] S21. Determine the phase arrangement entropy of the ocean background noise signal and the propeller cavitation noise signal in several channels of the same voyage at several scales within the same sampling period;

[0032] S22, determining the average value of the phase arrangement entropy corresponding to the ocean background noise signal and the propeller cavitation noise signal at the same scale in the same channel;

[0033] S23, determining the change curve of the phase arrangement entropy of the ocean background noise signal and the propeller cavitation noise signal in each channel with the scale factor according to the average value of the phase arrangement entropy in each channel;

[0034] S24. Determine the differentiated characteristics of the ocean background noise signal and the propeller cavitation noise signal according to the change curve, and realize nonlinear characteristic detection of underwater noise.

[0035] Furthermore, in step S1, when the information entropy is the improved permutation entropy, the method for extracting the improved permutation entropy is specifically:

[0036] B1. Normalizing the time series of underwater noise signals and performing coarse-graining on the normalized series to obtain a time subsequence;

[0037] B2, reconstruct the phase space of the time subsequence to obtain the phase space;

[0038] B3. Symbolize several row vectors of the phase space to obtain a symbolic phase space;

[0039] B4. According to the dimension of the reconstructed phase space, each group of row vectors in the symbol phase space is sorted to obtain the corresponding symbol pattern, and the probability of occurrence of each symbol pattern is determined;

[0040] B5. Determine the improved permutation entropy based on the probability of occurrence of each symbol pattern.

[0041] Furthermore, in step B1, the time series is normalized by using a cumulative distribution function, and the expression of the cumulative distribution function is:

[0042]

[0043] In the formula, a i is the original time series y i The normalized time series element, t is the normalized cumulative distribution factor, σ and μ are the mean and variance of the time series A, A = {a1, a2, a3, ..., a W}, subscript W is the total number of elements in time series A;

[0044] In step B1, the formula for coarse-graining the normalized sequence is:

[0045]

[0046] In the formula, is the element in the normalized sequence under scale s, s is the scale factor, and subscript j is the ordinal number of the element in the sequence after coarse-graining processing;

[0047] In step B2, the time subsequence y under scale s is (s) Reconstruct the phase space and get:

[0048]

[0049] Where Y i s is the phase space Y s The row vectors in is the row vector Y i s The elements in , τ is the time delay, and m is the dimension of the reconstructed phase space;

[0050] In step B3, the method for symbolizing several row vectors of the phase space is:

[0051] Phase space Y s The first row vector Y s (:,1)Symbolize through the averaging operator to obtain the symbolic phase space S s The first row vector S s (:,1):

[0052] Among them, the expression of the uniform quantization operator is:

[0053]

[0054] Where u is Y s The elements in (:,1), L is the preset discretization parameter, and is the phase space Y s Neutron sequence y s The minimum and maximum values ​​of , Δ is a discrete interval, and

[0055] Phase space Y s The k-th column and row vector Y s (:, k), 2≤k≤m, the corresponding symbolic result S is obtained by the following formula s (:,k);

[0056]

[0057] In the formula, S S (j,k) is the symbolic phase space S S The element in row j and column k, Y s (j,k) is the phase space Y s The k-th row and j-th column element in is the floor rounding symbol;

[0058] In step B5, the improved permutation entropy MIPE s The expression of (m,L) is:

[0059]

[0060] In the formula, p l is the probability of occurrence of symbol pattern number l, subscript 1≤l≤L m .

[0061] Furthermore, when the probability of the symbol pattern has only one element as 1 and the rest as 0, the improved permutation entropy MIPE s (m,L) takes the minimum value 0;

[0062] When the symbol patterns are uniformly distributed, the improved permutation entropy MIPE s (m,L) takes the maximum value L m , then normalize it and get:

[0063]

[0064] Furthermore, the step S2 is specifically as follows:

[0065] X21. Determine the improved permutation entropy of the ocean background noise signal and propeller cavitation noise signal in several channels of the same voyage at several scales within the same sampling period;

[0066] X22, determine the average value of the improved permutation entropy corresponding to the ocean background noise signal and the propeller cavitation noise signal at the same scale in the same channel;

[0067] X23. According to the average value of the improved permutation entropy in each channel, determine the variation curve of the improved permutation entropy of the ocean background noise signal and the propeller cavitation noise signal in each channel with the scale factor;

[0068] X24. According to the change curve, the differentiated characteristics of the ocean background noise signal and the propeller cavitation noise signal are determined to realize the nonlinear characteristic detection of underwater noise.

[0069] The beneficial effects of the present invention are:

[0070] (1) The present invention utilizes two information entropy algorithms to extract and classify the features of cavitation noise and ocean background noise. Through feature extraction and analysis of measured data, it is proved that both multi-scale phase permutation entropy and improved multi-scale permutation entropy can well classify the features of the two types of underwater acoustic signals and realize nonlinear feature detection of underwater noise.

[0071] (2) The present invention realizes the characteristic differentiation of two underwater noise signals. It is proved through the multi-scale phase permutation entropy algorithm that when the scale factor is greater than 100, the phase permutation entropy value of the background noise gradually widens the gap with the phase permutation entropy value of the cavitation noise as the scale factor increases, indicating that the phase complexity of cavitation noise is higher than the phase complexity of ocean background noise. However, this characteristic only begins to emerge when the scale factor reaches a certain range. The phase permutation entropy value of cavitation noise is close to 1, with high complexity, while the phase permutation entropy value of ocean background noise is significantly less than 1 at high scales, with low complexity. The improved multi-scale permutation entropy algorithm can feature cavitation noise and ocean background noise at any scale within the selected scale factor value range. The improved permutation entropy value of ocean background noise is close to 1, with high complexity, while the improved permutation entropy value of cavitation noise is significantly less than 1, with low complexity.

[0072] (3) The present invention reveals the internal laws of the complexity of underwater acoustic signals through in-depth analysis of the nonlinear characteristics of cavitation noise and ocean background noise based on information entropy, and explores new methods for detecting and processing low signal-to-noise ratio underwater acoustic signals in complex ocean environments. In view of the characteristics of cavitation noise and ocean environmental noise, multi-scale phase permutation entropy and improved multi-scale permutation entropy are used to extract complexity features, realize the characteristic distinction between cavitation noise and ocean background noise, and improve the detection capability of passive sonar systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 This is a flow chart of the nonlinear characteristic detection method for underwater noise provided by the present invention.

[0074] Figure 2 A schematic diagram of the sequence mode provided by the present invention.

[0075] Figure 3 This is a schematic diagram of the change curve of the multi-scale phase arrangement entropy provided by the present invention.

[0076] Figure 4 This is a schematic diagram of the change curve of the multi-scale improved permutation entropy provided by the present invention. DETAILED DESCRIPTION

[0077] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0078] Embodiment 1:

[0079] A method for detecting nonlinear characteristics of underwater noise comprises the following steps:

[0080] S1, collecting underwater noise signals and extracting information entropy;

[0081] S2. Perform multi-scale feature classification on the permutation entropy to obtain differentiated features of underwater noise and realize nonlinear feature detection of underwater noise;

[0082] The underwater noise signal is an ocean background noise signal and a propeller cavitation noise signal of the same voyage, and the information entropy is a phase arrangement entropy or an improved arrangement entropy.

[0083] In one embodiment of the present invention, when the information entropy of the underwater noise signal is the phase permutation entropy, the multi-scale phase permutation entropy algorithm is used in this embodiment to detect the nonlinear characteristics of underwater noise; wherein, in the above step S1, the permutation entropy describes the uncertainty of another variable under the condition of a known variable, or the amount of information that can be provided. Based on its calculation principle, the permutation entropy can be used to analyze the nonlinear correlation between two variables. In this embodiment, the multi-scale phase permutation entropy algorithm combines the permutation entropy of the signal phase information with the coarse-graining technology. Based on this, in step S1, the method for extracting the phase permutation entropy is:

[0084] A1. Determine the instantaneous phase sequence of underwater noise signal;

[0085] A2, coarse-graining the instantaneous phase sequence to obtain an instantaneous phase subsequence;

[0086] A3, reconstructing the instantaneous phase subsequence in phase space to obtain several row vectors;

[0087] A4. Arrange the elements in each row vector in ascending order according to the dimension of the reconstructed phase space to obtain several order patterns;

[0088] A5. Determine the probability of each order pattern appearing in the phase space;

[0089] A6. Determine the phase arrangement entropy based on the probability of occurrence of each order pattern.

[0090] In step A2 of this embodiment, the formula for coarse-graining the instantaneous phase sequence is:

[0091]

[0092] In the formula, is the instantaneous phase subsequence y under scale s (s) The elements in , s is the scale factor, x i is the instantaneous phase of the signal, x i ∈X, X is the instantaneous phase sequence, X={x1,x2,x3,...,x N}, N is the total number of instantaneous phases in X, subscript i is the instantaneous phase sequence number, and subscript j is the instantaneous phase subsequence y formed after the coarse-grained processing of sequence X (s) The ordinal number of the element in;

[0093] The process in step A2 is also called the coarse-graining process, which is essentially the averaging of s data in the window. In fact, the coarse-graining process also corresponds to the low-frequency part of the Harr wavelet.

[0094] In the above step A3, the row vector Y i s for:

[0095]

[0096] In the formula, is the row vector Y i s The elements in , τ is the time delay, and m is the dimension of the reconstructed phase space;

[0097] In step A4, the order pattern obtained by arranging the elements in the row vector in ascending order is:

[0098]

[0099] Where k1, k2, ..., k m is the original position of each element, π i =[k1,k2,...,km ],π i is a symbol vector;

[0100] Among them, π i =[k1,k2,...,k m ] and the row vector Y i s A one-to-one correspondence is established, which actually represents the row vector For example, for the vector [1.1, 1.2, 1.3], its symbol vector is [1, 2, 3]. It should be noted that when there are identical elements in a vector, its arrangement is determined by the order in which the elements appear. For the vector [1.1, 1.2, 1.3], its symbol vector is still [1, 2, 3]. It is easy to see that for an m-dimensional row vector, there may be m! Different arrangements, each of which is called an ordinal pattern (OP). When m = 3, the schematic diagrams of the 6 different OPs are as follows Figure 2 As shown;

[0101] Therefore, in the above step A5, the probability P of each order pattern appearing is l for:

[0102] P l =h l / N-m+1

[0103] In the formula, h l is the frequency of occurrence of each order pattern, subscript l = 1, 2, ..., m!;

[0104] In the above step A6, the phase arrangement entropy PE(m) is defined by the Shannon entropy, and its expression is:

[0105]

[0106] Step S2 in this embodiment is specifically as follows:

[0107] S21. Determine the phase arrangement entropy of the ocean background noise signal and the propeller cavitation noise signal in several channels of the same voyage at several scales within the same sampling period;

[0108] S22, determining the average value of the phase arrangement entropy corresponding to the ocean background noise signal and the propeller cavitation noise signal at the same scale in the same channel;

[0109] S23, determining the change curve of the phase arrangement entropy of the ocean background noise signal and the propeller cavitation noise signal in each channel with the scale factor according to the average value of the phase arrangement entropy in each channel;

[0110] S24. Determine the differentiated characteristics of the ocean background noise signal and the propeller cavitation noise signal according to the change curve, and realize nonlinear characteristic detection of underwater noise.

[0111] In this embodiment, the phase permutation entropy feature extraction is performed on the measured cavitation noise and ocean background noise of the same voyage. The voyage has nine channels with thousands of sampling cycles of collected data, of which the cavitation noise and ocean background noise data have 100 cycles each. We extract the same 10 sampling cycles of data at equal intervals in each channel as samples, set the scale to 1-150, calculate the phase permutation entropy of each period of data at each scale, and average the entropy values ​​of the same channel at the same scale for 10 periods to obtain the change curve of the phase permutation entropy of the two types of data samples in 9 channels with the scale factor as follows Figure 3 As shown in the figure, it can be seen that the phase permutation entropy of propeller cavitation noise and ocean background noise is almost indistinguishable when the scale factor is less than 100. When the scale factor is greater than 100, the phase permutation entropy of ocean background noise gradually distances itself from the phase permutation entropy of cavitation noise as the scale factor increases. This phenomenon shows that the phase complexity of cavitation noise is higher than that of ocean background noise, but this characteristic only begins to emerge when the scale factor reaches a certain range. The phase permutation entropy value of propeller cavitation noise is close to 1, and the complexity is high. The phase permutation entropy value of ocean background noise is significantly less than 1 at high scales, and the complexity is low.

[0112] In one embodiment of the present invention, improved permutation entropy extraction is performed by improving the multiscale permutation entropy MIPE (Multiscale Improved Permutation Entropy) algorithm. MIPE is a combination of the improved permutation entropy algorithm IPE (Improved Permutation Entropy, IPE) based on a single scale and the coarse-graining technology. The algorithm can improve the traditional permutation entropy without considering the signal phase information, improper processing of embedded equal elements, and other problems, and better suppress the interference of noise.

[0113] Based on this, in step S1 of this embodiment, when the information entropy of the underwater noise signal is the improved permutation entropy, the method for extracting the improved permutation entropy is specifically as follows:

[0114] B1. Normalizing the time series of underwater noise signals and performing coarse-graining on the normalized series to obtain a time subsequence;

[0115] B2, reconstruct the phase space of the time subsequence to obtain the phase space;

[0116] B3. Symbolize several row vectors of the phase space to obtain a symbolic phase space;

[0117] B4. According to the dimension of the reconstructed phase space, each group of row vectors in the symbol phase space is sorted to obtain the corresponding symbol pattern, and the probability of occurrence of each symbol pattern is determined;

[0118] B5. Determine the improved permutation entropy based on the probability of occurrence of each symbol pattern.

[0119] In the above step B1, the time series is normalized by using the cumulative distribution function, and the expression of the cumulative distribution function is:

[0120]

[0121] In the formula, a i is the original time series y i The normalized time series element, t is the normalized cumulative distribution factor, σ and μ are the mean and variance of the time series A, A = {a1, a2, a3, ..., a W}, subscript W is the total number of elements in time series A;

[0122] In the above step B1, the formula for coarse-graining the normalized sequence is:

[0123]

[0124] In the formula, is the element in the normalized sequence under scale s, s is the scale factor, and subscript j is the ordinal number of the element in the sequence after coarse-graining processing;

[0125] In step B2, the time subsequence y under scale s is (s) Reconstruct the phase space and get:

[0126]

[0127] In the formula, is the phase space Y s The row vectors in is the row vector Y i s The elements in , τ is the time delay, and m is the dimension of the reconstructed phase space;

[0128] In the above step B3, the method for symbolizing several row vectors in the phase space is:

[0129] Phase space Y s The first row vector Y s (:,1)Symbolize through the averaging operator to obtain the symbolic phase space S s The first row vector S s (:,1):

[0130] Among them, the expression of the uniform quantization operator is:

[0131]

[0132] Where u is Y s The elements in (:,1), L is the preset discretization parameter, and is the phase space Y s Neutron sequence y s The minimum and maximum values ​​of , Δ is a discrete interval, and

[0133] Phase space Y s The k-th column and row vector Y s (:, k), 2≤k≤m, the corresponding symbolic result S is obtained by the following formula s (:,k);

[0134]

[0135] In the formula, S S (j,k) is the symbolic phase space S S The element in row j and column k, Y s (j,k) is the phase space Y s The k-th row and j-th column element in is the floor rounding symbol;

[0136] In the above steps B4-B5, similar to the above steps A4-A6, the MIPE algorithm in this embodiment identifies each row in the symbol phase space as a "mode" π l In this embodiment, the “mode” in the MIPE algorithm is represented by the symbol pattern (Symbol Pattern, SP), and the probability p of each SP appearing in the statistical symbol phase space is l In step B5 of this embodiment, the improved permutation entropy is finally defined by the Shannon entropy, which is expressed as:

[0137]

[0138] In the formula, p l is the probability of occurrence of symbol pattern number l, subscript 1≤l≤L m ;

[0139] It should be noted that when the probability of the symbol pattern has only one element as 1 and the rest as 0, the improved permutation entropy MIPE s (m,L) takes the minimum value 0; when the symbol pattern obeys a uniform distribution, the improved permutation entropy MIPE s (m,L) takes the maximum value L m , then normalize it and get:

[0140]

[0141] Step S2 of this embodiment is specifically as follows:

[0142] X21. Determine the improved permutation entropy of the ocean background noise signal and propeller cavitation noise signal in several channels of the same voyage at several scales within the same sampling period;

[0143] X22, determine the average value of the improved permutation entropy corresponding to the ocean background noise signal and the propeller cavitation noise signal at the same scale in the same channel;

[0144] X23. According to the average value of the improved permutation entropy in each channel, determine the variation curve of the improved permutation entropy of the ocean background noise signal and the propeller cavitation noise signal in each channel with the scale factor;

[0145] X24. According to the change curve, the differentiated characteristics of the ocean background noise signal and the propeller cavitation noise signal are determined to realize the nonlinear characteristic detection of underwater noise.

[0146] In this embodiment, the same data samples as in the previous embodiment are used, and the scales are set to 1-40. The improved permutation entropy of each periodic data at each scale is calculated. The entropy values ​​of 10 periods of the same channel at the same scale are summed to obtain the curve of the improved permutation entropy of the two types of data samples of 9 channels with the scale factor. Figure 4 As shown in the figure, it can be seen that within the selected scale range, the improved permutation entropy can feature classify cavitation noise and ocean background noise. The improved permutation entropy value of ocean background noise is close to 1, with high complexity, and the improved permutation entropy value of cavitation noise is significantly less than 1, with low complexity.

[0147] Embodiment 3:

[0148] In this embodiment, the measured propeller cavitation noise and ocean background noise data of the same voyage are used. The voyage consists of 9 channels with thousands of sampling cycles of collected data, of which the propeller cavitation noise and ocean background noise data each have 100 cycles. The same 10 sampling cycles of data are collected at equal intervals in each channel as samples, and the scale is set to 1 to 40. The phase permutation entropy and improved permutation entropy at each scale of each period of data are calculated, and the entropy values ​​of 10 periods of the same channel at the same scale are averaged to obtain the change curves of the two types of data samples of 9 channels with the phase permutation entropy and the improved permutation entropy with the scale factor as the classification basis of the two types of underwater acoustic signals.

[0149] like Figure 3 , 4 As shown in the figure, the phase permutation entropy and improved permutation entropy of propeller cavitation noise and ocean background noise vary with the scale factor. Figure 3 It can be seen from the figure that when the scale factor increases to a certain range, the phase complexity of cavitation noise and ocean background noise begins to be clearly distinguished; Figure 4 It can be seen that at any scale, the improved permutation entropy of the two types of underwater acoustic signals shows obvious differences. In addition, the entropy algorithm can only perform stable and accurate entropy estimation when the length of the data involved in the calculation meets certain requirements, while the coarse-grained technology decreases the length of the coarse-grained sequence as the scale factor increases. When the amount of data is small, the multi-scale phase permutation entropy may no longer be applicable. The improved multi-scale permutation entropy can effectively classify the two types of underwater acoustic signals at any low scale, and the application is more robust.

[0150] The above data analysis proves that both algorithms can effectively extract the characteristic differences between cavitation noise and ocean background noise and achieve effective classification. This result has important reference value for further detection and classification of underwater target signals.

Claims

1. A method for detecting nonlinear characteristics of underwater noise, characterized in that: The following steps are involved: S1, collecting underwater noise signals and extracting information entropy; S2, performing multi-scale feature classification on the permutation entropy to obtain differentiated features of underwater noise and realize nonlinear feature detection of underwater noise; Wherein, the underwater noise signal is the ocean background noise signal and propeller cavitation noise signal of the same voyage, and the information entropy is the phase arrangement entropy; In step S1, the method for extracting phase arrangement entropy is: A1. Determine the instantaneous phase sequence of the underwater noise signal; A2, coarse-graining the instantaneous phase sequence to obtain an instantaneous phase subsequence; A3, reconstructing the instantaneous phase subsequence in phase space to obtain several row vectors; A4. Arrange the elements in each row vector in ascending order according to the dimension of the reconstructed phase space to obtain several order patterns; A5. Determine the probability of each order pattern appearing in the phase space; A6. Determine the phase arrangement entropy based on the probability of occurrence of each order pattern; The step S2 is specifically as follows: S21. Determine the phase arrangement entropy of the ocean background noise signal and the propeller cavitation noise signal in several channels of the same voyage at several scales within the same sampling period; S22, determining the average value of the phase arrangement entropy corresponding to the ocean background noise signal and the propeller cavitation noise signal at the same scale in the same channel; S23, determining the change curve of the phase arrangement entropy of the ocean background noise signal and the propeller cavitation noise signal in each channel with the scale factor according to the average value of the phase arrangement entropy in each channel; S24. Determine the differentiated characteristics of the ocean background noise signal and the propeller cavitation noise signal according to the change curve, and realize nonlinear characteristic detection of underwater noise.

2. The nonlinear characteristic detection method of underwater noise according to claim 1, characterized in that: In step A2, the formula for coarse-graining the instantaneous phase sequence is: In the formula, For scale The instantaneous phase subsequence The elements in is the scale factor, is the instantaneous phase of the signal, , is the instantaneous phase sequence, , for The total number of instantaneous phases in is the instantaneous phase sequence, subscript is the instantaneous phase subsequence formed after the coarse-grained processing of sequence X The ordinal number of the element in; In step A3, the row vector for: In the formula, is a row vector The elements in For time delay, is the dimension of the reconstructed phase space; In step A4, the order pattern obtained by arranging the elements in the row vector in ascending order is: In the formula, is the original position of each element. , is a symbolic vector; In step A5, the probability of each order pattern appearing is for: In the formula, is the frequency of occurrence of each order pattern, subscript ; In step A6, the phase arrangement entropy The expression is: 。

Images