A method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD

Abstract

This invention discloses a method for suppressing self-noise signals from sonar buoys based on DCCA and EEMD, belonging to the field of underwater acoustic signal processing technology. The method includes the following steps: collecting acoustic data and motion wave velocity data for sonar buoys affected by self-noise interference under sea wave excitation; calculating broadband correlation coefficients using DCCA; calculating IMF component correlation coefficients using EEMD and DCCA; comparing the magnitudes of the IMF component correlation coefficients and broadband correlation coefficients to perform fuzzy filtering of self-noise; calculating narrowband self-noise data; further identifying noisy data using the narrowband self-noise data; and reconstructing the output by combining all noise-free data. This invention starts from the mechanism of sonar buoy self-noise generation, deeply explores the inherent correlation between sonar buoy self-noise and sea wave interference, mines the cumulative probability density distribution law of cross-correlation coefficients, and constructs a sonar buoy self-noise suppression method based on correlation criteria under sea wave excitation, achieving effective suppression of self-noise.

Description

Technical Field

[0001] This invention relates to the field of underwater acoustic signal processing technology, and in particular to a method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD. Background Technology

[0002] Sonar buoys are air-dropped acoustic devices specifically designed for underwater target detection. They consist of an underwater acoustic receiver and a surface transmitter. These devices are capable of searching for, locating, and identifying underwater targets, making them ideal for large-scale joint detection. However, in practical applications, sonar buoys are susceptible to self-noise interference, affecting their detection performance. To this day, sonar buoys remain a crucial means of underwater target detection, but effectively suppressing self-noise remains a key area of ​​research and improvement.

[0003] Suppressing sonar buoy self-noise interference can be achieved through two methods: front-end structural optimization and back-end signal processing. Self-noise mainly originates from the sonar buoy's own mechanical vibration and the influence of sea surface waves on the hydrophone. The former optimizes the sonar buoy structure to decouple the surface and underwater units, thereby suppressing the vertical movement of the hydrophone caused by sea wave fluctuations and reducing the resulting self-noise. The latter, based on the signal characteristics of sonar buoy self-noise, performs noise reduction processing on the acoustic signals received by the hydrophone to eliminate or reduce the impact of self-noise, thus improving detection accuracy. Currently, research on sonar buoy self-noise suppression based on structural optimization is relatively complete, while research on suppressing sonar buoy self-noise using signal processing methods is relatively weak and requires further in-depth study.

[0004] Based on the above, a method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD, so as to solve the problems in the background art.

[0006] To achieve the above objectives, this invention provides a method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD, comprising the following steps:

[0007] S1. Acoustic data collection is performed on sonar buoys that are affected by self-noise under the excitation of sea surface waves to obtain the raw signal and record the vertical motion wave velocity data of the sonar buoy in the corresponding time domain.

[0008] S2. The DCCA algorithm is used to perform cross-correlation analysis on the original signal and the vertical motion wave velocity data of the sonar buoy to obtain the broadband correlation coefficient under the condition of broadband self-noise data.

[0009] S3. The EEMD algorithm is used to perform mode decomposition on the original signal to obtain the IMF component. The DCCA algorithm is used to perform cross-correlation analysis on the IMF component and the sonar buoy vertical motion wave velocity data to obtain the correlation coefficient of the IMF component.

[0010] S4. Compare the correlation coefficients of the IMF components and the broadband correlation coefficients. If the correlation coefficient of the IMF components is lower than that of the broadband correlation coefficient, fuzzy filter out the data without self-noise interference; otherwise, output the remaining data containing self-noise interference.

[0011] S5. Reconstruct and perform short-time Fourier transform on the data containing self-noise interference output by S4 to obtain narrowband self-noise data at different frequencies.

[0012] S6. Further identify and remove noisy data using narrowband self-noise data. Perform inverse short-time Fourier transform on the remaining noise-free data and reconstruct it in conjunction with the data without self-noise interference in S4 to output a noise-reduced signal, thus completing the suppression of the sonar buoy's self-noise signal.

[0013] Preferably, in step S2, the specific process of performing cross-correlation analysis using the DCCA algorithm is as follows:

[0014] S21. Accumulate and calculate the original signal and the vertical motion wave velocity data of the sonar buoy to obtain the corresponding accumulated signal;

[0015] S22. Divide the two accumulated signals obtained in S21 into Nn overlapping windows, each window containing n+1 values;

[0016] Where N is the sequence length and n is the number of overlapping elements;

[0017] S23. Within each window, the corresponding local trend is fitted using the least squares method.

[0018] S24. Within each window, eliminate local trends to obtain residual sequences, and calculate the mean of the detrended covariance values ​​of all residual sequences.

[0019] S25. Based on the results of S24, calculate the DCCA cross-correlation coefficient as the broadband correlation coefficient.

[0020] Preferably, in step S23, the specific calculation process for fitting using the least squares method is as follows:

[0021] 1) Construct a polynomial linear regression model, which is expressed as:

[0022]

[0023] In the formula, Provides the position index for each element within the window. The coefficients of the multinomial regression model are... The value of the signal accumulated within the window;

[0024] 2) Simplify the polynomial linear regression model. The simplified polynomial linear regression model is expressed as:

[0025]

[0026] In the formula, I is the design matrix, which consists of a polynomial of position indices within the window. Y represents the polynomial coefficients. S The value of the signal accumulated within the window;

[0027] 2) Calculate the polynomial coefficient estimates for both sequences. The polynomial coefficient estimates are expressed as follows:

[0028]

[0029] In the formula, These are estimates of the polynomial coefficients of sequence X. Pinv(I) is the polynomial coefficient estimate of sequence Y, and Pinv(I) is the generalized inverse matrix of matrix I.

[0030] 3) Calculate the local trends separately. The local trends are represented as follows:

[0031]

[0032] In the formula, Let X be the local trend of sequence X in the i-th window. Let Y be the local trend of sequence Y in the i-th window.

[0033] Preferably, in S25, the DCCA cross-correlation coefficient is expressed as:

[0034]

[0035] In the formula, ρ DCCA Let F be the cross-correlation coefficient of DCCA. DCCA (n) is the mean of the detrended covariance values, Fn XDFA (n) is the mean of the detrended covariance between sequence X and its own sequence, F YDFA (n) is the mean of the detrended covariance between sequence Y and its own sequence.

[0036] Preferably, the detrended covariance value is expressed as:

[0037]

[0038] In the formula, fDCCA (n,i) represents the detrended covariance, i is the starting index of the window, k is the index within the current window, and R0 k R' is the accumulated signal of sequence X. k The accumulated signal of sequence Y, Let X be the local trend of sequence X in the i-th window. Let Y be the local trend of sequence Y in the i-th window.

[0039] Preferably, in step S3, the specific process of using the EEMD algorithm for mode decomposition is as follows:

[0040] S31. Add P equal-length and finite-amplitude Gaussian white noise to the original signal obtained in S1 to obtain the initial signal;

[0041] S32. Perform EMD decomposition on the initial signal to obtain multiple IMFs and margins;

[0042] S33. Take the average value of all IMFs to obtain the IMF components.

[0043] Preferably, in step S32, the specific process of EMD decomposition is as follows:

[0044] 1) Identify local extrema in the initial signal. Based on the local extrema, use the spline difference function to fit the upper and lower envelopes of the initial signal. Calculate the average value of the upper and lower envelopes and regard the difference between the initial signal and the average value as the intermediate signal.

[0045] 2) Determine whether the intermediate signal meets the IMF requirements. If it does, use the intermediate signal as the first IMF. If it does not, use the intermediate signal as the initial signal and repeat step 1) until a new intermediate signal meets the IMF requirements and is used as the first IMF.

[0046] 3) Separate the first IMF from the initial signal, and repeat steps 1) to 2) with the separated signal as the new initial signal to obtain the second IMF. Continue in this manner until the m-th IMF is obtained. When the separated signal becomes monotonic, the EMD decomposition is completed.

[0047] Preferably, in step S5, the narrowband self-noise data at different frequencies is represented as follows:

[0048]

[0049] In the formula, STFT x (n,k) represents narrowband self-noise data at different frequencies, x(n+m) represents the signal points of the input signal within the current window, ω(m) is the window function, x(n+m)ω(m) is a short-time sequence, and e -j2πnk / NIt is a complex exponential function, and j is the imaginary unit.

[0050] Preferably, in step S6, the specific process for identifying noisy data is as follows:

[0051] 1) The DCCA algorithm was used to perform cross-correlation analysis on the narrowband self-noise data and the vertical motion wave velocity data of the sonar buoy in S5 to obtain the narrowband correlation coefficient;

[0052] 2) If the narrowband correlation coefficient is higher than the broadband correlation coefficient, the data is noisy and should be removed; otherwise, the data is retained and output as noise-free data.

[0053] Therefore, the sonar buoy self-noise signal suppression method based on DCCA and EEMD of the present invention has the following beneficial effects:

[0054] (1) By leveraging the advantages of the DCCA algorithm in processing cross-correlation of non-stationary time series data, the correlation distribution law between waves and self-noise was effectively explored, and the inherent correlation characteristics were extracted, providing a quantitative analysis method for sonar buoy self-noise suppression.

[0055] (2) By combining the adaptive decomposition capability of EEMD and the DCCA algorithm to effectively extract the self-noise correlation characteristics of sonar buoys, and conducting in-depth research on the correlation matching mechanism between data, a sonar buoy self-noise suppression method based on correlation criteria was constructed. This method can further suppress the self-noise of sonar buoys on the basis of optimizing the sonar front-end structure, thereby achieving effective suppression of self-noise and improving the sonar buoy's ability to search, locate, identify and detect underwater targets.

[0056] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0057] Figure 1 This is a general framework diagram of an embodiment of the present invention;

[0058] Figure 2 This is an EEMD set empirical mode decomposition diagram according to an embodiment of the present invention;

[0059] Figure 3 This is a time-frequency diagram of self-noise data according to an embodiment of the present invention;

[0060] Figure 4 The original signal, noise reduction data, and vertical motion wave velocity diagram of the sonar buoy are shown in the embodiments of the present invention.

[0061] Figure 5 This is a time-frequency diagram of the noise reduction data in an embodiment of the present invention;

[0062] Figure 6This is a spectrum diagram of the original signal and self-noise data in an embodiment of the present invention;

[0063] Figure 7 This is a spectrum diagram of the original signal and the noise-reduced data in an embodiment of the present invention. Detailed Implementation

[0064] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0066] Example

[0067] like Figure 1 As shown, the following simulation test is conducted using a sonar buoy self-noise signal suppression method based on DCCA and EEMD provided by this invention to verify the suppression effect of the correlation criterion on the sonar buoy self-noise signal. The steps are as follows:

[0068] S1. Acoustic data collection is performed on sonar buoys that are affected by self-noise under the excitation of sea surface waves to obtain the raw signal and record the vertical motion wave velocity data of the sonar buoy in the corresponding time domain.

[0069] S2. The DCCA algorithm is used to perform cross-correlation analysis on the original signal obtained in S1 and the vertical motion wave velocity data of the sonar buoy to obtain the broadband correlation coefficient under broadband self-noise data conditions.

[0070] DCCA (Detrending Cross-Correlation Analysis) integrates the advantages of correspondence analysis and multiple regression analysis. It can not only detect the autocorrelation of non-stationary time series, but also effectively eliminate trend components of various orders in the signal. It can effectively analyze the long-range correlation of non-stationary and nonlinear time series.

[0071] The specific process of performing cross-correlation analysis using the DCCA algorithm is as follows:

[0072] S21. Perform cumulative calculations on the two sequences separately (in S2, when using the DCCA algorithm, the two sequences are the original signal and the sonar buoy vertical motion wave velocity data, respectively; in S3, when using the DCCA algorithm, the two sequences are the IMF component and the sonar buoy vertical motion wave velocity data, respectively; in S6, when identifying noisy data, cross-correlation analysis is performed on the two sequences of narrowband self-noise data and sonar buoy vertical motion wave velocity data), to obtain the corresponding cumulative signal, specifically:

[0073] Sequences X and Y are represented as follows:

[0074] X = {x(i), i = 1, ..., N};

[0075] Y = {y(i), i = 1, ..., N};

[0076] The accumulated signals of the two are as follows:

[0077]

[0078] In the formula, R k R' is the accumulated signal of sequence X. k This is the accumulated signal of sequence Y;

[0079] S22. Divide the two accumulated signals obtained in S21 into Nn overlapping windows. Each window contains n+1 values, and the values ​​of each window start from i and end at i+n.

[0080] Where N is the sequence length and n is the number of overlapping elements;

[0081] S23. Within each window in S22, the corresponding local trend is fitted using the least squares method.

[0082] The specific calculation process for fitting using the least squares method is as follows:

[0083] 1) Construct a polynomial linear regression model, which is expressed as:

[0084]

[0085] In the formula, Provides the position index for each element within the window. The coefficients of the multinomial regression model are... The value of the signal accumulated within the window;

[0086] The above formula can be simplified to:

[0087]

[0088] In the formula, I is the design matrix, which consists of a polynomial of position indices within the window. Y represents the polynomial coefficients. S The value of the signal accumulated within the window;

[0089] 2) For sequences X and Y, we have:

[0090]

[0091] In the formula, These are estimates of the polynomial coefficients of sequence X. Pinv(I) is the polynomial coefficient estimate of sequence Y, and Pinv(I) is the generalized inverse matrix of matrix I.

[0092] 3) Calculate local trends:

[0093]

[0094] In the formula, Let X be the local trend of sequence X in the i-th window. This represents the local trend of sequence Y in the i-th window;

[0095] S24. Within each window in S22, eliminate the local trend obtained in S23 to obtain the residual sequence, and calculate the mean of the detrended covariance values ​​of all residual sequences.

[0096] The detrended covariance is expressed as:

[0097]

[0098] In the formula, f DCCA (n,i) represents the detrended covariance, i is the starting index of the window, k is the index within the current window, and R0 k R' is the accumulated signal of sequence X. k This is the accumulated signal of sequence Y;

[0099] The mean of the detrended covariance is expressed as:

[0100]

[0101] In the formula, F DCCA (n) is the mean of the detrended covariance values, Fn DCCA (n) is approximately a power function with n as the base and DCCA scaling parameter λ as the exponent. The magnitude of DCCA scaling parameter λ can reflect the cross-correlation between two sequences.

[0102] If 0 < λ < 0.5, it indicates a long-range negative correlation between the two sequences; if 0.5 < λ < 1, it indicates a long-range positive correlation between the two sequences. However, the DCCA scaling parameter cannot strictly quantify the degree of correlation between the two sequences, so the DCCA cross-correlation coefficient needs to be introduced to quantify the degree of correlation between the sequences.

[0103] S25. Based on the results of S24, calculate the DCCA cross-correlation coefficient;

[0104] The cross-correlation coefficient of DCCA is expressed as:

[0105]

[0106] In the formula, ρ DCCALet F be the cross-correlation coefficient of DCCA. DCCA (n) is the mean of the detrended covariance values, Fn XDFA (n) is the mean of the detrended covariance between sequence X and its own sequence, F YDFA (n) is the mean of the detrended covariance between sequence Y and its own sequence;

[0107] The magnitude of the DCCA cross-correlation coefficient can reflect the cross-correlation between sequences: if ρ DCCA =0 indicates that there is no cross-correlation between the two sequences; if 0 < ρ DCCA <1 indicates a positive correlation between the two sequences, and when ρ DCCA When ρ = 1, the two sequences are strictly positively correlated; if -1 < ρ DCCA <0 indicates a negative correlation between the two sequences, and when ρ DCCA When =-1, the two sequences are strictly negatively correlated.

[0108] S3. The EEMD algorithm is used to perform mode decomposition on the original signal obtained in S1 to obtain the IMF component. The DCCA algorithm in S2 is used to perform cross-correlation analysis on the IMF component and the sonar buoy vertical motion wave velocity data obtained in S1 to obtain the correlation coefficient of the IMF component.

[0109] The EEMD (Empirical Mode Decomposition) method requires no prior data and has no restrictions on the type of signal. Compared with traditional methods such as least squares fitting, moving average, and low-pass filtering, it exhibits better feature extraction capabilities in processing complex non-stationary and nonlinear time series data.

[0110] like Figure 2 As shown, the specific process of mode decomposition using the EEMD algorithm is as follows:

[0111] S31. Add P equal-length and finite-amplitude Gaussian white noise to the original signal obtained in S1 to obtain the initial signal;

[0112] The initial signal is represented as:

[0113] x i (n)=x(n)+n i (n);

[0114] In the formula, x(n) is the original signal, and n i (n) represents the Gaussian white noise added for the i-th time, x i (n) is the signal after the i-th addition of Gaussian white noise, i.e., the initial signal;

[0115] S32. Perform EMD decomposition on the initial signal to obtain multiple IMFs and margins, specifically:

[0116] 1) Identify local extrema in the initial signal. Based on the local extrema, use the spline difference function to fit the upper and lower envelopes of the initial signal. Calculate the average value of the upper and lower envelopes and regard the difference between the initial signal and the average value as the intermediate signal.

[0117] The average value of the upper and lower envelopes is expressed as:

[0118]

[0119] In the formula, m(n) is the average value of the upper envelope and the lower envelope, y(t) is the upper envelope, and l(t) is the lower envelope;

[0120] The intermediate signal is represented as:

[0121] h(n) = x i (n)-m(n);

[0122] In the formula, h(n) is the intermediate signal;

[0123] 2) Determine whether the intermediate signal meets the IMF requirements. If it does, use the intermediate signal as the first IMF. If it does not, use the intermediate signal as the initial signal and repeat step 1) until a new intermediate signal meets the IMF requirements and is used as the first IMF.

[0124] The IMF requires that the difference between the number of extreme points and the number of zero points does not exceed 1, and that the average of the upper and lower envelopes is zero.

[0125] 3) Separate the first IMF from the initial signal, and repeat steps 1) to 2) with the separated signal as the new initial signal to obtain the second IMF. Continue in this manner until the m-th IMF is obtained. When the separated signal becomes monotonic, the EMD decomposition is completed.

[0126] After adding Gaussian white noise multiple times and performing EMD decomposition, the set of initial signals is represented as:

[0127]

[0128] In the formula, C i,j (n) represents the j-th order IMF obtained after adding white noise for the i-th time, r i (n) represents the residual obtained after the i-th addition of white noise;

[0129] S33. Take the average value of all IMFs to obtain the IMF components;

[0130] The IMF component is represented as:

[0131]

[0132] In the formula, c j (n) represents the IMF component finally obtained after the EEMD algorithm;

[0133] After decomposition by the EEMD algorithm, the original signal is represented as:

[0134]

[0135] In the formula, M is the number of IMFs obtained from EEMD decomposition, and r m (n) represents the residual components after EEMD decomposition. It can be seen that the decomposition result on the right-hand side implicitly contains an assumption, namely, x... i (n) The number of IMF components obtained after each decomposition is the same. However, in actual decomposition, the number of IMFs may vary, which introduces new problems for ensemble averaging. Furthermore, after each addition of Gaussian white noise, x... i (n) is decomposed independently, and each decomposition yields a remainder r. i (n), that is:

[0136]

[0137] Therefore, when using the EEMD method to analyze signals, the reconstructed signals after decomposition still contain self-noise that could not be eliminated after ensemble averaging, and further screening of the self-noise that could not be eliminated is required.

[0138] S4. Compare the IMF component correlation coefficient obtained in S3 with the broadband correlation coefficient obtained in S2. If the IMF component correlation coefficient is lower than the broadband correlation coefficient, fuzzy filter out the data without self-noise interference; otherwise, output the remaining data containing self-noise interference.

[0139] S5. Reconstruct and perform short-time Fourier transform on the data containing self-noise interference output by S4 to obtain narrowband self-noise data at different frequencies.

[0140] like Figure 3 As shown, the narrowband self-noise data at different frequencies are represented as follows:

[0141]

[0142] In the formula, STFT x (n,k) represents the spectral value at the k-th frequency within the n-th time window, reflecting the narrowband self-noise data at different frequencies. x(n+m) represents the signal point of the input signal within the current window, ω(m) is the window function, and x(n+m)ω(m) is the short-time sequence. -j2πnk / N It is a complex exponential function, and j is the imaginary unit.

[0143] S6, such as Figure 4 As shown, the narrowband self-noise data obtained in S5 is used to further identify and remove noisy mode data. The remaining noiseless mode data is then subjected to inverse short-time Fourier transform and reconstructed in conjunction with the data in S4 that does not contain self-noise interference. The noise-reduced signal is then output to complete the suppression of the sonar buoy's self-noise signal.

[0144] The specific process for identifying noisy data is as follows:

[0145] 1) The DCCA algorithm in S2 is used to perform cross-correlation analysis on the narrowband self-noise data in S5 and the sonar buoy vertical motion wave velocity data obtained in S1 to obtain the narrowband correlation coefficient.

[0146] 2) If the narrowband correlation coefficient is higher than the broadband correlation coefficient in S2, it is noisy data and should be removed; otherwise, it should be retained and noisy data should be output.

[0147] like Figure 6 As shown, this embodiment also collects and analyzes acoustic data for sonar buoys that are only affected by self-noise and are not excited by sea surface waves. It calculates the waveform, frequency, and sound pressure level of the self-noise and obtains the sonar number, time, and data of the self-noise, which are used to analyze the self-noise characteristics of the sonar buoy.

[0148] The formula for calculating the self-noise frequency is:

[0149]

[0150] In the formula, X[k] is the self-noise frequency, and DFT(x[n]) represents the discrete Fourier transform of the self-noise time-domain signal. As a weighting factor;

[0151] The weighting factor is expressed as:

[0152]

[0153] like Figure 5 , Figure 7 As shown, the self-noise frequency is compared with the noise-reduced signal output after self-noise suppression processing and the original signal data. It can be seen that the correlation criterion can effectively improve the suppression effect on the self-noise signal of sonar buoys. In addition to being applied to the self-noise suppression of sonar buoys, this method can also be applied to the suppression of acoustic array current noise. It can be used in conjunction with active sonar to improve the accuracy and efficiency of target detection and has good application prospects.

[0154] Therefore, this invention provides a sonar buoy self-noise signal suppression method based on DCCA and EEMD. By leveraging the advantages of the DCCA algorithm in processing cross-correlation of non-stationary time series data, it effectively explores the correlation distribution law between waves and self-noise and extracts the inherent correlation characteristics, providing a quantitative analysis method for sonar buoy self-noise suppression.

[0155] By combining the adaptive decomposition capability of EEMD and the DCCA algorithm to effectively extract the self-noise correlation characteristics of sonar buoys, and by studying the correlation matching mechanism between data, a sonar buoy self-noise suppression method based on correlation criteria was constructed. This method can further suppress sonar buoy self-noise on the basis of optimizing the sonar front-end structure, thereby achieving effective suppression of self-noise and improving the sonar buoy's ability to search, locate, and identify underwater targets.

[0156] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD, characterized in that, Includes the following steps: S1. Acoustic data collection is performed on sonar buoys that are affected by self-noise under the excitation of sea surface waves to obtain the raw signal and record the vertical motion wave velocity data of the sonar buoy in the corresponding time domain. S2. The DCCA algorithm is used to perform cross-correlation analysis on the original signal and the vertical motion wave velocity data of the sonar buoy to obtain the broadband correlation coefficient under the condition of broadband self-noise data. S3. The EEMD algorithm is used to perform mode decomposition on the original signal to obtain the IMF component. The DCCA algorithm is used to perform cross-correlation analysis on the IMF component and the sonar buoy vertical motion wave velocity data to obtain the correlation coefficient of the IMF component. S4. Compare the correlation coefficients of the IMF components and the broadband correlation coefficients. If the correlation coefficient of the IMF components is lower than that of the broadband correlation coefficient, fuzzy filter out the data without self-noise interference; otherwise, output the remaining data containing self-noise interference. S5. Reconstruct and perform short-time Fourier transform on the data containing self-noise interference output by S4 to obtain narrowband self-noise data at different frequencies. S6. Further identify and remove noisy data using narrowband self-noise data. Perform inverse short-time Fourier transform on the remaining noise-free data and reconstruct it in conjunction with the data without self-noise interference in S4 to output a noise-reduced signal, thus completing the suppression of the sonar buoy's self-noise signal.

2. The method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD according to claim 1, characterized in that, In S2, the specific process of using the DCCA algorithm for cross-correlation analysis is as follows: S21. Accumulate and calculate the original signal and the vertical motion wave velocity data of the sonar buoy to obtain the corresponding accumulated signal; S22. Divide the two accumulated signals obtained in S21 into Nn overlapping windows, each window containing n+1 values; Where N is the sequence length and n is the number of overlapping elements; S23. Within each window, the corresponding local trend is fitted using the least squares method. S24. Within each window, eliminate local trends to obtain residual sequences, and calculate the mean of the detrended covariance values ​​of all residual sequences. S25. Based on the results of S24, calculate the DCCA cross-correlation coefficient as the broadband correlation coefficient.

3. The method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD according to claim 2, characterized in that, In S25, the DCCA cross-correlation coefficient is expressed as: In the formula, ρ DCCA Let F be the cross-correlation coefficient of DCCA. DCCA (n) is the mean of the detrended covariance values, Fn XDFA (n) is the mean of the detrended covariance between sequence X and its own sequence, f YDFA (n) is the mean of the detrended covariance between sequence Y and its own sequence.

4. The method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD according to claim 3, characterized in that, The detrended covariance value is expressed as: In the formula, f DCCA (n, i) represents the detrended covariance value, i is the starting index of the window, k is the index within the current window, and R0 k Let R′ be the accumulated signal of sequence X. k The accumulated signal of sequence Y, Let X be the local trend of sequence X in the i-th window. Let Y be the local trend of sequence Y in the i-th window.

5. The method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD according to claim 1, characterized in that, In S3, the specific process of using the EEMD algorithm for mode decomposition is as follows: S31. Add P equal-length and finite-amplitude Gaussian white noise to the original signal obtained in S1 to obtain the initial signal; S32. Perform EMD decomposition on the initial signal to obtain multiple IMFs and margins; S33. Take the average value of all IMFs to obtain the IMF components.

6. The method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD according to claim 5, characterized in that, In S32, the specific process of EMD decomposition is as follows: 1) Identify local extrema in the initial signal. Based on the local extrema, use the spline difference function to fit the upper and lower envelopes of the initial signal. Calculate the average value of the upper and lower envelopes and regard the difference between the initial signal and the average value as the intermediate signal. 2) Determine whether the intermediate signal meets the IMF requirements. If it does, use the intermediate signal as the first IMF. If it does not, use the intermediate signal as the initial signal and repeat step 1) until a new intermediate signal meets the IMF requirements and is used as the first IMF. 3) Separate the first IMF from the initial signal, and repeat steps 1) to 2) with the separated signal as the new initial signal to obtain the second IMF. Continue in this manner until the m-th IMF is obtained. When the separated signal becomes monotonic, the EMD decomposition is completed.

7. The method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD according to claim 1, characterized in that, In S5, the narrowband self-noise data at different frequencies are represented as follows: In the formula, STFT x (n, k) represents narrowband self-noise data at different frequencies, x(n+m) represents the signal points of the input signal within the current window, x(n+m)ω(m) represents the short-time sequence, and e -j2πnk / N It is a complex exponential function, and j is the imaginary unit.

8. A method for suppressing self-noise signals of sonar buoys based on DCCA and EEMD according to claim 1, characterized in that, In step S6, the specific process for identifying noisy data is as follows: 1) The DCCA algorithm was used to perform cross-correlation analysis on the narrowband self-noise data and the vertical motion wave velocity data of the sonar buoy in S5 to obtain the narrowband correlation coefficient; 2) If the narrowband correlation coefficient is higher than the broadband correlation coefficient, the data is noisy and should be removed; otherwise, the data is retained and output as noise-free data.

Images