Underwater target shaft frequency electric field signal comprehensive detection method

Abstract

The invention discloses a comprehensive detection method for an underwater target shaft frequency electric field signal, which comprises the following steps of: detecting a target signal by using a target signal detection algorithm based on the combination of EEMD (Ensemble Empirical Mode Decomposition) and improved power spectrum entropy, and performing line spectrum feature enhancement on the detected target signal through three-stable stochastic resonance and an LMP (Line Management Protocol) algorithm. According to the method, false detection caused by blind line spectrum feature enhancement on a pseudo signal is avoided, the prominence and detectability of the line spectrum feature of the target signal are improved, the method has good target identification capability, the signal processing efficiency of the algorithm is improved, and the method is suitable for popularization and application. Functional complementation and technical collaboration of target signal detection and line spectrum feature enhancement are realized. According to the invention, systematic detection functions of target signal detection, line spectrum feature enhancement, identification and intensity value extraction are realized on the whole, and the method has good engineering practicability.

Description

Technical Field

[0001] The present invention belongs to the technical field of signal detection, and particularly relates to a comprehensive detection method for the shaft-frequency electric field signal of an underwater target. Background Art

[0002] The existing underwater target detection technologies mainly rely on physical field signals such as sound field, electric field, and magnetic field. Among them, although the sound field signal has a good propagation effect at a relatively long distance, with the development of shock absorption and noise reduction technologies, the technical difficulties and challenges faced by existing sonar devices in the process of target detection and positioning are gradually increasing. Compared with the sound field signal, the detection of the electric field signal has some obvious advantages. Especially in complex water area conditions, the electric field signal has strong stability and a certain resistance to the influence of hydrological and meteorological changes. Therefore, it has important engineering application value and popularization prospects to use the shaft-frequency electric field signal generated during the movement of a ship at sea for target detection.

[0003] The shaft-frequency electric field signal is a weak signal with low-frequency and line-spectrum characteristics, mainly originating from the rotation of the ship's propeller. The fundamental frequency of the shaft-frequency electric field signal is usually between 0.5 Hz and 7 Hz. Due to its significant line-spectrum characteristics, the shaft-frequency electric field signal has strong stability and detectability, and is suitable for target detection and recognition. Although the advantages of using the shaft-frequency electric field signal for target detection and recognition are obvious, there are still great technical challenges in the current related technical fields. The amplitude of the shaft-frequency electric field signal itself is small, and the signal is easily interfered by the electric field environmental noise, resulting in difficulty in accurately extracting and recognizing the signal in a low signal-to-noise ratio environment. Moreover, the spectrum analysis ability of the traditional spectrum analysis method based on the fast Fourier transform (FFT) is insufficient. When dealing with weak signals, these traditional spectrum analysis methods are difficult to obtain accurate frequency components in the detection of shaft-frequency signals in the low-frequency band. In addition, there are also problems of low detection accuracy and insufficient real-time performance in the existing ship shaft-frequency electric field detection process. Thus, it can be seen that the current related technical fields still lack and need to improve the dedicated detection algorithms tailored for the shaft-frequency electric field signal.

[0004] Regarding the characteristics that the shaft-frequency electric field signal is mainly concentrated in the low-frequency band and has obvious line-spectrum characteristics, Ji Dou et al. proposed a signal processing algorithm based on tristable stochastic resonance. This algorithm can effectively enhance weak periodic signals, which is beneficial to enhancing the target shaft-frequency electric field signal. However, this algorithm is sensitive to noise types and signal parameters, and requires relatively complex parameter adjustment during use. Improper use is likely to enhance false signals, resulting in misjudgment. Huang Yong et al. also proposed the least mean P-norm (LMP) algorithm, which can effectively suppress broadband noise, further highlighting the line-spectrum characteristics of the shaft-frequency electric field signal for target detection. However, in a strong interference or low signal-to-noise ratio environment, the enhancement effect of this algorithm is unstable, and it is easy to mis-enhance the interference spectral lines in false signals, thus affecting the reliability of target detection. Zhang Jiawei et al. proposed an algorithm for detecting shaft-frequency electric field signals based on ensemble empirical mode decomposition (EEMD) and improved power spectral entropy. This algorithm can accurately detect weak target signals, but it cannot actively enhance the line-spectrum characteristics of target signals, which is not conducive to subsequent target recognition using the line-spectrum characteristics of target signals in a strong noise background. Summary of the Invention

[0005] To overcome the deficiencies of the prior art, the present invention provides a comprehensive detection method for underwater target shaft-frequency electric field signals. First, a target signal detection algorithm based on the combination of EEMD and improved power spectral entropy is used to detect the target signal, and then the detected target signal is subjected to line-spectrum feature enhancement through tristable stochastic resonance and LMP algorithms. This not only avoids blindly enhancing the line-spectrum features of false signals, resulting in false detection, but also improves the prominence and detectability of the line-spectrum features of target signals. It not only has good target recognition ability, but also improves the signal processing efficiency of the algorithm, realizing the complementarity of functions and the technical cooperation between target signal detection and line-spectrum feature enhancement. Overall, the present invention realizes the systematic detection functions of detecting, enhancing line-spectrum features, recognizing, and extracting intensity values of target signals, and has good engineering practicability.

[0006] The technical solutions adopted by the present invention to solve its technical problems are as follows:

[0007] Step 1: Signal preprocessing;

[0008] Step 2: Target signal detection;

[0009] Step 3: Use the tristable stochastic resonance method to enhance the line-spectrum features of the target signal;

[0010] Step 4: Use the LMP algorithm to enhance the line-spectrum features of the target signal after tristable stochastic resonance;

[0011] Step 5: Extract the characteristic frequency of the target signal and perform target recognition.

[0012] Preferably, step 1 is specifically as follows:

[0013] Preprocess the collected electric field signals as follows:

[0014] Select the electric field signals in the frequency band of 0.5 Hz to 8 Hz and perform band-pass filtering to obtain band-pass filtered signals ;

[0015] Use the moving average method of a sliding window with a length of to smooth the signals and thus obtain the final preprocessed signals ;

[0016] Preferably, step 2 is specifically as follows:

[0017] Step 2-1: Using as the detection window duration and as the data window duration, perform signal detection using a sliding window at the current moment , and intercept the data of the previous points at the current moment from the preprocessed signals as the signal detection sequence , where is the sampling frequency of the electric field signal ;

[0018] Step 2-2: Use the EEMD method to decompose the signal detection sequence into multiple intrinsic mode function IMF components, and estimate the power spectra of the first IMF components decomposed; then, after equally dividing the entire frequency axis according to the number of frequency bands , obtain frequency sub-intervals ;

[0019] Step 2-3: Based on the calculation method of the energy peak entropy ratio EPER eigenvalue, calculate the EPER eigenvalue for each of the frequency sub-intervals divided, and the calculation formula is defined as follows:

[0020] (1)

[0021] In the formula: is the serial number of each frequency sub-interval ; is the power spectrum value of the th frequency point in each frequency sub-interval ; is the frequency sub - interval mid - frequency point power spectrum maximum value; is the maximum - value operation; is the number of frequency points in each frequency sub - interval; represents the frequency sub - interval normalized power spectrum information entropy;

[0022] Step 2 - 4: Use the K - means clustering algorithm to classify frequency sub - intervals according to the EPER eigenvalue of each frequency sub - interval ;

[0023] The K - means clustering algorithm is implemented based on the objective function of the minimum sum of squared errors within a class ; The calculation formula of

[0024] (2)

[0025] In the formula: is the maximum number of clusters; is the cluster number; is the sequence number of each frequency sub - interval ; is the cluster; is the frequency sub - interval EPER eigenvalue; is the cluster the average value of the EPER eigenvalues of the frequency sub - intervals in it; The symbol represents the relationship that an element belongs to a set; The symbol is to find the minimum value of the data; represents and the squared error;

[0026] Step 2 - 5: Select the cluster with the largest EPER eigenvalue in the clusters of the frequency sub - intervals through the K - means clustering algorithm as the set of alternative frequency sub - intervals , discard the frequency sub - intervals that continuously appear in the same IMF, and then use the remaining alternative sub - intervals with the largest EPER eigenvalue in as the target signal frequency sub - intervals , according to The detection feature quantity required for signal detection is calculated from the data therein , and the calculation formula of the detection feature quantity is defined as:

[0027] (3)

[0028] In the formula: is the detection feature quantity; is the current detection moment of the sliding window; G is the number of frequency points in the target signal frequency sub-interval ; is the target signal frequency sub-interval The power spectrum values at each frequency point within; is the target signal frequency sub-interval The frequency point index within;

[0029] Step 2-6: When the detection feature quantity is greater than the dynamic threshold , it is determined that the target signal is detected at the current detection moment ;

[0030] The dynamic threshold is calculated using a method combining the mean and standard deviation, and the calculation formula is defined as:

[0031] (4)

[0032] In the formula: and are respectively the average value and standard deviation of the detection feature quantity in the previous moments before the current detection moment ; is the threshold factor; is the weight factor that controls the weight of the standard deviation in the calculation of the threshold value;

[0033] The calculation formula of is defined as:

[0034] (5)

[0035] In the formula: is the sliding step of the sliding window;

[0036] The current detection moment before The average value of the detection feature quantity in the previous moments

[0037] (6)

[0038] In the formula: is the detection feature value extracted from the sliding window at the

[0039] current detection time before detection feature quantities in the standard deviation of is defined by the calculation formula:

[0040] (7)

[0041] Step 2-7: For the preprocessed signal execute Steps 2-1 to 2-6 three times in a loop. If the target signal is detected two or more times in Step 2-6 during these three times, output the final detection decision result as the target signal is detected; otherwise, output that the target is not detected.

[0042] Preferably, the specific content of Step 3 is as follows:

[0043] For the preprocessed signal with the detected target shaft frequency electric field signal characteristics , perform line spectrum feature enhancement processing using the bistable stochastic resonance method. The calculation formula is as follows:

[0044] (8)

[0045] In the formula: is the target signal after bistable stochastic resonance processing at time; is the nonlinear coefficient of the system.

[0046] Preferably, the specific content of Step 4 is as follows:

[0047] Use the LMP algorithm to continue processing the target signal after the bistable stochastic resonance processing method. The calculation formula is defined as follows:

[0048] (9)

[0049] In the formula: is the output response at time after being processed by the LMP algorithm, is the FIR adaptive filter weight coefficient vector at time, represents the transpose operation on the target signal .

[0050] Preferably, the specific content of Step 5 is as follows:

[0051] Step 5-1: By means of complex exponential projection spectrum analysis, extract the characteristic frequency and its corresponding amplitude of the target signal processed by the LMP algorithm ; By performing complex exponential projection on each frequency point of the target signal within a frequency range, find the frequency point with the maximum complex exponential projection value as the characteristic frequency of the target signal , and according to the corresponding projection modulus length inversely deduce its corresponding amplitude ; The calculation formula of the maximum complex exponential projection value is defined as:

[0052] (10)

[0053] In the formula: is the total number of sample points of the target signal ; is the value of the target signal at the th sampling point; is the time corresponding to the th sampling point; is the characteristic frequency of the target signal; is the complex exponential basis function of the characteristic frequency of the target signal; [[ID=??]]

[0054] Step 5-2: The calculation formula of the estimated amplitude corresponding to the characteristic frequency of the target signal in the frequency domain is defined as:

[0055] (11)

[0056] In the formula: is the modulo operation;

[0057] Step 5-3: By comparing the extracted characteristic frequency of the target signal with the existing database of shaft frequency ranges of different ship types, the type of the target ship can be identified.

[0058] A computer program that causes a computer to execute the above comprehensive detection method.

[0059] An electronic device, comprising: a processor and a memory; the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory so that the electronic device executes the above comprehensive detection method.

[0060] A computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, the above comprehensive detection method is implemented.

[0061] A chip includes: a processor for calling and running a computer program from a memory, so that a device equipped with the chip executes the above comprehensive detection method.

[0062] The beneficial effects of the present invention are as follows:

[0063] (1) Signal detection and signal enhancement are strictly separated to avoid introducing false signals;

[0064] The method of the present invention directly performs detection on the original signal, avoiding the introduction of false components by the bistable stochastic resonance and the LMP algorithm in the early stage of signal processing. Since the bistable stochastic resonance and the LMP belong to non-linear processing algorithms, although they can enhance the signal, they may also amplify the noise or generate false spectral lines. If enhanced at the beginning, it is easy to misdetect. Detect first and then enhance to ensure the reliability of the front-end detection, which can effectively reduce the misdetection rate and better meet the application requirements in actual engineering.

[0065] (2) Targeted enhancement after detection, with a smaller algorithm calculation amount;

[0066] The method of the present invention does not require enhancing the entire long-time data by the bistable stochastic resonance and the LMP algorithm. It only needs to enhance the data segments that have been detected to contain the line spectrum features of the target signal line, greatly reducing the unnecessary calculation amount and being very suitable for target detection platforms with limited real-time processing power consumption.

[0067] (3) The combination of the bistable stochastic resonance and the LMP algorithm has complementary advantages;

[0068] The bistable stochastic resonance of the method of the present invention mainly improves the signal-to-noise ratio of weak signals and is particularly suitable for signals with obvious line spectrum features such as shaft frequency signals. It can mainly significantly enhance the line spectrum features of the target signal line contained in the noisy signal. The LMP algorithm can further significantly enhance the line spectrum features of the known target frequency and adaptively filter out the noise. In the process of enhancing the target signal, in the first step, the line spectrum features of the target signal line are roughly enhanced by the bistable stochastic resonance, so that the line spectrum features of the target signal line are as obvious as possible. In the second step, the LMP algorithm is used to further accurately enhance the enhanced line spectrum features of the target signal line. This two-stage enhancement strategy of rough enhancement plus fine enhancement significantly improves the algorithm processing effect compared with using only one algorithm for signal processing.

[0069] (4) Suitable for low signal-to-noise ratio and weak feature environments;

[0070] The method of the present invention first realizes accurate detection of the signal through EEMD and improved power spectrum entropy, avoiding blind processing of the entire noisy signal, making the signal processing more concentrated on the part that truly contains the target signal, and enabling subsequent algorithms to separate the signal and noise more efficiently. After the detection of the target signal is completed, the local noisy signal containing the target component is processed by the tristable stochastic resonance and the LMP algorithm, making full use of the amplification ability of the tristable stochastic resonance for weak periodic signals and the advantages of the LMP algorithm in non-Gaussian noise suppression, significantly improving the resolvability of the shaft spectral line features and the target recognition rate, so that the present invention is applicable to the target detection task in an environment with low signal-to-noise ratio and weak features.

[0071] (5) Facilitate subsequent high-level processing such as target positioning;

[0072] Since the method of the present invention only enhances the signal line spectrum features of the original noisy signal for which the target signal has been detected, it improves the accuracy of signal detection. And because the characteristic frequency and signal intensity value of the target signal are effectively extracted, it is conducive to subsequent operations such as target positioning, and is very suitable for ship electric field detection projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 It is a schematic diagram of the overall architecture of the method of the present invention;

[0074] Figure 2 It is a schematic diagram of the signal preprocessing process in the method of the present invention;

[0075] Figure 3 It is a schematic diagram of the target signal detection process in the method of the present invention;

[0076] Figure 4 It is a schematic diagram of the tristable stochastic resonance algorithm process in the method of the present invention;

[0077] Figure 5 It is a schematic diagram of the least mean P-norm (LMP) algorithm process in the method of the present invention;

[0078] Figure 6 It is a schematic diagram of the target recognition process in the method of the present invention;

[0079] Figure 7 It is the simulation time-domain waveform and simulation spectrogram of the electric field signal received by the sensor in the embodiment of the present invention;

[0080] Figure 8 It is the simulation time-domain waveform and simulation spectrogram of the preprocessed signal in the embodiment of the present invention;

[0081] Figure 9 It is the simulation time-domain waveform and simulation spectrogram of the first IMF component in the embodiment of the present invention;

[0082] Figure 10 This is the simulation result diagram of the target signal's tristable stochastic resonance in the embodiment of the present invention;

[0083] Figure 11 This is the simulation result diagram of the processing of the target signal by the minimum average P-norm (LMP) algorithm in the embodiment of the present invention. Specific implementation manners

[0084] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0085] Aiming at the problems existing in the existing underwater target shaft-frequency electric field signal detection technology, such as limited ability to identify target signals, blindness in the process of enhancing line spectrum features, low computational efficiency in the algorithm processing process, and poor adaptability to real-time detection, the present invention proposes a comprehensive detection method for underwater target shaft-frequency electric field signals. Based on the target signal algorithm processing strategy of "detect first and then enhance", through the target signal detection algorithm combining EEMD and improved power spectrum entropy, the target signal is first detected and locked, and then the tristable stochastic resonance and LMP algorithm are used to specifically enhance the line spectrum features of the detected target signal, thereby highlighting its line spectrum features, effectively improving the detectability and identifiability of the target signal. The present invention realizes the efficient complementarity of the detection and enhancement functions, effectively improves the operation efficiency of the algorithm while enhancing the target recognition accuracy, and has good system integration and application promotion value.

[0086] The present invention provides a comprehensive detection method for underwater target shaft-frequency electric field signals. The overall architecture schematic diagram of this method is as shown in the attached Figure 1 figure, and the specific technical solutions adopted by the present invention are divided into the following steps:

[0087] Step 1: Signal preprocessing;

[0088] The flow schematic diagram of step 1-1: signal preprocessing is as shown in Figure 2 the figure. A Butterworth band-pass filter with an order of is used to intercept the signal in the 0.5 Hz - 8 Hz frequency band of the electric field signal and perform band-pass filtering to obtain the signal . The calculation definition formula of the electric field signal is:

[0089] (12)

[0090] In the formula: is the time vector; is the amplitude intensity of the target signal; is the noise signal; represents the sine function symbol;

[0091] Time vector The calculation formula is defined as:

[0092] (13)

[0093] Where: represents a discrete sequence;

[0094] Step 1-2: Use a moving average method with a sliding window of length to smooth the signal to obtain a preprocessed signal .

[0095] Step 2: Detect the target signal based on an algorithm combining EEMD and improved power entropy;

[0096] Step 2-1: The flow diagram of target signal detection is as shown in Figure 3 . At time , use a sliding window to detect the preprocessed signal , and intercept the points of data before the current time as the signal detection sequence . .

[0097] Step 2-2: Decompose into multiple IMF components by EEMD, and then estimate the power spectra of the first IMF components respectively, and divide the entire frequency axis into equal-length frequency sub-intervals according to the number of frequency bands to obtain frequency sub-intervals . The calculation formula for the power spectrum of the IMF component is defined as:

[0098] (14)

[0099] Where: is the index of the frequency domain sampling point; is the index of the IMF component; is the imaginary unit; is the index of the time domain sampling point; is the th power spectrum value of the th IMF component at the th frequency point; is the th sampling point in the time domain of the th IMF component; is the operation of taking the modulus square of a complex number. is a complex exponential function, which represents the phase and amplitude of each frequency component in the frequency domain;

[0100] Step 2-3: Calculate the divided Frequency subintervals EPER eigenvalue . The calculation formula is defined as:

[0101] (15)

[0102] Step 2-4: Use K-means clustering method to select frequency subintervals The class with the largest EPER eigenvalue among the clustered As a set of alternative frequency subintervals , the K-means clustering method is mainly based on the objective function of the minimum square error sum within the class Perform frequency subinterval classification, The calculation formula is defined as:

[0103] (16)

[0104] Step 2-5: Discard the alternative frequency sub-intervals The frequency subintervals that appear continuously in the same IMF component and are Among the remaining frequency subintervals, the frequency subinterval with the largest EPER eigenvalue is selected as the target signal frequency subinterval ,pass The data in can be used to calculate the current time Detection feature quantity in the lower sliding window , detection feature quantity The calculation formula is defined as:

[0105] (17)

[0106] Step 2-6: Use historical detection features in the current sliding window The mean and standard deviation Calculating dynamic thresholds Dynamic Threshold The calculation formula is defined as:

[0107] (18)

[0108] If the defined continuous detection time The detection feature quantity calculated in the sliding window is Greater than its corresponding dynamic threshold , then the signal detection sequence is considered It is a valid signal, indicating the presence of the target signal.

[0109] Step 2-7: For the preprocessed signal Repeat the processing of Step 2-1 to Step 2-6, with a total of three loop repetitions, and perform a voting decision on the final signal detection result based on the three signal detection results in Step 2-6. That is, if the target signal is detected two or more times in the three signal detections, it is comprehensively judged that the target signal is detected; otherwise, it is judged that the target signal is not found.

[0110] Step 3: Use bistable stochastic resonance to enhance the line spectrum characteristics of the target signal;

[0111] Step 3-1: After comprehensively judging that the target signal is detected, use the bistable stochastic resonance algorithm to process the preprocessed signal to obtain the target signal after signal processing , and this algorithm can improve the signal-to-noise ratio of the target signal , thereby enhancing the line spectrum characteristics of the target signal . The flow schematic diagram of the bistable stochastic resonance algorithm is as shown in Figure 4 . In the present invention, the first-order bistable stochastic resonance algorithm is used to enhance the line spectrum characteristics of the target signal. The calculation formula of this algorithm is defined as:

[0112] (19)

[0113] Step 3-2: Use the fourth-order Runge-Kutta method to numerically solve the differential equation shown in Equation (18), and then the target signal after being processed by the bistable stochastic resonance algorithm for the preprocessed signal can be obtained. The calculation formula for numerically approximating and solving Equation (18) using the fourth-order Runge-Kutta method can be defined as:

[0114] (20)

[0115] In the formula: and are the approximate solution values of the signal at the moment and the moment respectively; , , , are different slope estimation values calculated within the time step .

[0116] The calculation formula of

[0117] ​ (21)

[0118] The calculation formula of

[0119] (22)

[0120] The calculation formula of

[0121] (23)

[0122] The calculation formula of

[0123] (24)

[0124] In equations (21), (22), (23), and (24): is the time step; and respectively represent the moment of and the preprocessed signal input at the moment of ; is the nonlinear coefficient of the system; and are respectively the moment of and the approximate solution value at the moment of

[0125] The differential equation (19) is numerically approximated step by step by iteratively updating the state quantity at each time step, and finally the output response signal of the triple-well stochastic resonance is obtained.

[0126] Step 4: Use the LMP algorithm to enhance the line spectrum characteristics of the target signal after triple-well stochastic resonance;

[0127] Step 4-1: Apply the LMP algorithm to the target signal after triple-well stochastic resonance for signal processing, so as to further highlight the line spectrum characteristics of the target signal. The flow schematic diagram of the LMP algorithm is as shown in Figure 5 , and the main process of the LMP algorithm for processing the target signal is as follows:

[0128] The estimation error of the LMP algorithm for the target signal at the moment of is defined as:

[0129] (25)

[0130] In equation (25): and are the weight coefficient vector and the input target signal vector of the FIR adaptive filter at moment respectively; is the expected response of the LMP algorithm system at denotes the transpose operation on the signal ;

[0131] In formula (25), the adjustment and update formula of the weight coefficient vector is:

[0132] (26)

[0133] In formula (26): is the sign function; is the step size factor of the algorithm; is the order of the algorithm; is the estimation error at denotes the input target signal at is the estimation error of the absolute power of order

[0134] Step 4-2: The output response of the target signal after being processed by the LMP algorithm at moment is:

[0135] (27)

[0136] By using the calculation formulas of each variable in the above steps 4-1 to 4-2, the LMP algorithm can finally complete the processing of the target signal by the LMP algorithm, so as to obtain the target signal after being processed by the LMP algorithm;

[0137] Step 5: Target signal characteristic frequency extraction and target recognition;

[0138] Step 5-1: The flow chart of target recognition is as shown in Figure 6 . By using the method of finding the maximum complex exponential projection value to extract the target signal characteristic frequency from the target signal after being processed by the LMP algorithm. The calculation formula of the maximum complex exponential projection value is defined as:

[0139] (28)

[0140] Step 5-2: Feature frequencies of the target signal in the frequency domain and corresponding amplitudes are extracted, and the amplitude is estimated The calculation formula is defined as:

[0141] (29)

[0142] Step 5-3: By comparing the extracted feature frequencies of the target signal with the shaft frequency characteristic ranges of various types of ships in Table 1, target recognition can be achieved.

[0143] Table 1 Shaft frequency ranges of different types of ships

[0144] Ship type Axial frequency range / Hz Oil tanker, large cargo ship 1.33~1.75 Fast regular cargo ship 2.00~2.33 Passenger ship warship 2.33~3.003.33~6.67

[0145] Each step of the present invention is closely linked and cooperates with each other. Through a processing flow from rough to fine, key information in the target frequency band is gradually extracted and strengthened, effectively suppressing interference from irrelevant noise, so as to ensure efficient and accurate signal detection and recognition even in a strong noise background.

[0146] Example:

[0147] In order to verify the effectiveness and reliability of the ship shaft frequency electric field detection method proposed by the present invention, an ideal simulation is used to test and experiment on the performance of the present invention. The specific process of the experimental simulation is as follows:

[0148] Steps for setting simulation parameters:

[0149] Step 1: Parameter setting of the electric field signal ;

[0150] (1) Feature frequency of the target signal Hz.

[0151] (2) Sampling frequency Hz.

[0152] (3) Time vector .

[0153] (4) Amplitude .

[0154] (5) The signal-to-noise ratio is calculated by intercepting the shaft frequency signal characteristic frequency band between 0.1 Hz and 7 Hz, and the specific value is .

[0155] Step 2: Parameter setting for signal preprocessing of ;

[0156] (1) The order of the Butterworth band-pass filter is , minimum cutoff frequency Hz, the highest cutoff frequency is Hz.

[0157] (2) The sliding window length of the sliding window moving average method is .

[0158] Step 3: Preprocess the signal Set parameters for target signal detection;

[0159] (1) Detection window duration s;

[0160] (2) Data window duration s;

[0161] (3) Sliding window sliding step s;

[0162] (4) Continuous detection duration s;

[0163] (5) Number of IMF components for signal decomposition by EEMD ;

[0164] (6) Number of frequency bands ;

[0165] (7) Number of categories clustered by the K-means clustering algorithm ;

[0166] (8) Threshold factor ;

[0167] (9) Standard deviation weighting coefficient .

[0168] Step 4: Preprocess the signal containing the target signal component Perform parameter setting for tristable stochastic resonance algorithm processing;

[0169] (1) Time step ;

[0170] (2) System parameters 、 、 .

[0171] Step 5: Target signal after tristable stochastic resonance processing Set the parameters for LMP algorithm processing;

[0172] (1) Adaptive filter order ;

[0173] (2)Adaptive step size ;

[0174] (3)Algorithm order 。

[0175] Step 6: Set the parameters for extracting the characteristic frequency and target recognition of the target signal after being processed by the LMP algorithm and target recognition;

[0176] (1)Search frequency range ;

[0177] (2)Frequency search step size 。

[0178] Steps for verifying and analyzing the simulation results:

[0179] Step 1: Add noise to the generated target source signal to simulate the electric field signal received by the electric field sensor during target detection in the ocean environment , and the time domain and frequency spectrum diagrams are as shown in Figure 7 . It can be seen from Figure 7 that the simulated electric field signal is submerged by noise in the time domain, and its line spectrum characteristics cannot be seen in the frequency domain. Therefore, it can simulate the signal received by the electric field sensor in the actual ocean environment with large noise interference.

[0180] Step 2: The preprocessed signal after band-pass filtering and smoothing operations on the electric field signal , and the time domain and frequency spectrum diagrams are as shown in Figure 8 . It can be seen from Figure 8 that the result of preprocessing the electric field signal meets the expectations.

[0181] Step 3: Detect the target signal from the preprocessed signal . EEMD decomposes the signal detection sequence into 5 IMF components, and the time domain and frequency spectrum diagrams of the first IMF component are as shown in Figure 9 . It can be known from the experimental simulation detection results that the target signal detection algorithm in the present invention has correctly detected the target signal.

[0182] Step 4: The simulation result diagram of the bistable stochastic resonance processing of the preprocessed signal judged to have target signal components after detection is as shown in Figure 10 . It can be seen from Figure 10 that after the preprocessed signal is processed by bistable stochastic resonance, the target signal , the spectral characteristics of the target signal line can already be observed more clearly in its spectrogram. At this time, the target signal has a signal-to-noise ratio of . Comparing with the signal-to-noise ratio of the original electric field signal , the signal-to-noise ratio has increased by . It can be seen that the tristable stochastic resonance algorithm effectively enhances the spectral characteristics of the target signal line.

[0183] Step 5: The simulation result diagram of processing the signal after tristable stochastic resonance processing again by the LMP algorithm is as shown in Figure 11 . It can already be clearly observed from Figure 11 the spectral characteristic components of the target signal. At this time, the target signal has a signal-to-noise ratio of . Comparing it with the signal-to-noise ratio of the target signal , the signal-to-noise ratio has increased by . It can be seen that on the basis of the tristable stochastic resonance processing of the noisy signal, the LMP algorithm successfully further highlights the spectral characteristics of the target signal and improves the recognizability of the target signal.

[0184] Step 6: Extract the characteristic frequency of the target signal after LMP algorithm processing. It can be known from the experimental simulation results that within a very small error range, the target signal characteristic frequency extraction algorithm used in the present invention has correctly extracted the target signal characteristic frequency Hz, and the signal amplitude corresponding to the extraction of the target signal characteristic frequency in the frequency domain at this point is 0.0041V. Comparing the target signal characteristic frequency with the corresponding data in Table 1 can achieve the accurate identification of the detection target.

[0185] In summary, through simulation verification and theoretical analysis, it is fully proved that the method proposed in the present invention has feasibility and promotion potential in the field of target detection of shaft-frequency electric field signals, and can achieve the correct detection, spectral characteristic enhancement, identification and classification, and amplitude intensity extraction of target shaft-frequency electric field signals, and has good application prospects.

Claims

1. An integrated detection method for the shaft-frequency electric field signal of an underwater target, characterized in that It includes the following steps: Step 1: Signal preprocessing; Step 2: Target signal detection; Step 3: Using the tristable stochastic resonance method to enhance the line spectrum features of the target signal; Step 4: Using the LMP algorithm to enhance the line spectrum features of the target signal after tristable stochastic resonance; Step 5: Target signal characteristic frequency extraction and target recognition.

2. The integrated detection method for the shaft-frequency electric field signal of an underwater target according to claim 1, wherein The specific content of Step 1 is as follows: For the collected electric field signals Perform preprocessing: Select the electric field signal in the frequency band of 0.5 Hz to 8 Hz Perform band-pass filtering to obtain a band-pass filtered signal ; The sliding window moving average method with a sliding window length of is used to perform data smoothing on the signal so as to obtain the final preprocessed signal .

3. The integrated detection method for the shaft-frequency electric field signal of an underwater target according to claim 2, wherein The specific content of Step 2 is as follows: Step 2-1: Using as the detection window duration, as the data window duration, at the current moment perform signal detection using a sliding window, and intercept the data of points before the current moment from the preprocessed signal as the signal detection sequence , where is the electric field signal sampling frequency; Step 2-2: Use the EEMD method to decompose the signal detection sequence into multiple intrinsic mode function IMF components, and estimate the power spectra of the first IMF components among the decomposed IMF components; then, after equally dividing the entire frequency axis according to the number of frequency bands , obtain frequency sub-intervals ; Step 2-3: Based on the calculation method of the energy peak entropy ratio (EPER) eigenvalue, calculate the EPER eigenvalue for each of the frequency sub-intervals one by one. The calculation formula is defined as follows: （1） In the formula: is the serial number of each frequency sub - interval ; is the power spectrum value of the th frequency point in each frequency sub - interval ; is the maximum power spectrum value of the th frequency point in the th frequency sub - interval ; is the operation of taking the maximum value; is the number of frequency points in each frequency sub - interval ; represents the normalized power spectrum information entropy within the th frequency sub - interval ; Step 2-4: Use the K-means clustering algorithm to classify the frequency sub-intervals according to the EPER eigenvalue of each frequency sub-interval ;​ The K-means clustering algorithm is based on the objective function of the within-class minimum sum of squared errors and is implemented The calculation formula of which is defined as: （2） Wherein: is the maximum number of clusters; is the serial number of the cluster; is each frequency sub - interval sequence number; is the th cluster; is the th frequency sub - interval EPER eigenvalue; is the th cluster in which the mean value of the EPER eigenvalues of each frequency sub - interval ; the symbol represents the relationship that an element belongs to a set; the symbol is to find the minimum value of data; represents and square error; Step 2-5: Select the cluster with the largest EPER eigenvalue among the clusters of frequency sub-intervals through the K-means clustering algorithm as the set of alternative frequency sub-intervals , discard the frequency sub-intervals that continuously appear in the same IMF in , and use the remaining alternative sub-interval with the largest EPER eigenvalue in as the target signal frequency sub-interval , calculate the detection feature quantity required for signal detection based on the data in , and the calculation formula of the detection feature quantity is defined as: （3） Wherein: is the detection feature quantity; is the current detection time of the sliding window; G is the number of frequency points within the target signal frequency sub-interval ; is the power spectrum value at each frequency point within the target signal frequency sub-interval ; is the frequency point index within the target signal frequency sub-interval ; Step 2-6: When the detected feature quantity is greater than the dynamic threshold at the current detection moment a target signal is detected; Dynamic threshold is calculated using a method that combines the mean and standard deviation. The calculation formula is defined as follows: （4） Wherein: and are respectively the average value and the standard deviation of the detected feature quantity in the moments before the current detection moment ; is the threshold factor; is the weight factor for controlling the weight of the standard deviation in the calculation of the threshold value; The calculation formula is defined as: （5） Wherein: is the sliding step length of the sliding window; Current detection time Previous Detection feature quantities in Average value The calculation formula is defined as: （6） In the formula: is the detection feature value extracted on the sliding window at the th moment; Current detection time Previous The standard deviation of the detected feature quantity in the previous time instants is defined by the following calculation formula: (7) Step 2-7: For the preprocessed signal Execute steps 2-1 to 2-6 three times in a loop. If the target signal is detected two or more times in step 2-6 during these three times, output the final detection decision result as the target signal is detected; otherwise, output that the target is not detected.

4. The integrated detection method for the axial frequency electric field signal of an underwater target according to claim 3, characterized in that The specific content of Step 3 is as follows: For the preprocessed signal detected with the characteristic of the target shaft-frequency electric field signal , the bistable stochastic resonance method is used for line-spectrum characteristic enhancement processing, and the calculation formula is as follows: （8） In the formula: is the target signal after tristable stochastic resonance processing at moment; is the nonlinear coefficient of the system.

5. The integrated detection method for the shaft-frequency electric field signal of an underwater target according to claim 4, characterized in that The specific content of Step 4 is as follows: The target signal after being processed by the tristable stochastic resonance processing method is further processed using the LMP algorithm, and the calculation formula is defined as follows: Continue to process, and the calculation formula is defined as follows: （9） In the formula: is the output response at time after being processed by the LMP algorithm, is the weight coefficient vector of the FIR adaptive filter at time denotes the transpose operation on the target signal ​ 6. The integrated detection method for the shaft-frequency electric field signal of an underwater target according to claim 5, wherein The specific content of Step 5 is as follows: Step 5-1: By means of complex exponential projection spectrum analysis, extract the characteristic frequency and its corresponding amplitude of the target signal processed by the LMP algorithm ; By performing complex exponential projection on each frequency point of the target signal within a frequency range, find the frequency point with the maximum complex exponential projection value as the characteristic frequency of the target signal , and according to the corresponding projection modulus inversely deduce its corresponding amplitude ; The calculation formula of the maximum complex exponential projection value is defined as: (10) Where: is the total number of sample points of the target signal ; is the value of the target signal at the -th sampling point; is the time corresponding to the -th sampling point; is the characteristic frequency of the target signal; is the complex exponential basis function of the characteristic frequency of the target signal; Step 5-2: Target signal characteristic frequency The estimated amplitude corresponding in the frequency domain is defined by the calculation formula as follows: （11） In the formula: is the modulo operation; Step 5-3: By comparing the extracted target signal characteristic frequencies with the existing database of shaft frequency ranges for different ship types, the type of the target ship can be identified.

7. A computer program, characterized in that, The computer program causes the computer to execute the method described in any one of claims 1 to 6.

8. An electronic device, characterized in that, It includes: A processor and a memory; The memory is used to store the computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device executes the method described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method described in any one of claims 1 to 6.

10. A chip, characterized in that, It includes: A processor, which is used to call and run the computer program from the memory, so that the device installed with the chip executes the method described in any one of claims 1 to 6.

Images