Ship shaft frequency electric field feature extraction method based on eigenmode function energy permutation entropy

Abstract

The invention discloses a ship shaft frequency electric field feature extraction method based on eigenmode function energy permutation entropy. The ship shaft frequency electric field feature extraction method comprises the following steps that original shaft frequency electric field signals, collected by a signal collection device, of a plurality of target objects are decomposed into eigenmode functions (IMF) through empirical mode decomposition (EMD); obtaining a plurality of intrinsic mode function IMF components; calculating instantaneous energy and average energy of each IMF component; sorting the IMF components according to the average energy value of the IMF components; selecting the IMF component with the maximum average energy from the sequenced IMF components; calculating a permutation entropy (PE) value of the selected maximum average energy IMF component; taking the PE value as a signal feature to perform extraction and target object distinguishing; different ship target signals can be effectively distinguished, and the problem of screening different ship signals is solved.

Description

Technical Field

[0001] The invention belongs to the field of feature extraction, and in particular relates to a method for extracting ship shaft frequency electric field features based on eigenmode function energy permutation entropy. Background Art

[0002] Due to the low frequency, slow attenuation, and rich frequency content of ship shaft-frequency electric field signals, they have become an important method for underwater target detection after acoustic field detection. However, effectively extracting target features in complex marine environments is a key research focus and hot topic in underwater target detection. Traditional signal feature extraction methods mainly include time-domain signal structure, Fourier transform of time-domain signals, classical spectrum estimation, and wavelet transform. Ship shaft-frequency electric field signals are non-stationary and nonlinear, making it difficult to directly reflect effective parameters of target features from the raw signal. The Fourier transform integrates the signal over the entire time domain and cannot clearly reflect the time-varying characteristics of non-stationary electrical signals. The wavelet transform, affected by the wavelet basis function, is prone to signal energy leakage, making it impossible to accurately extract time-domain signal features. In summary, traditional signal feature extraction methods are no longer able to solve the problem of feature extraction from complex shaft-frequency signals under low signal-to-noise ratio conditions. Summary of the Invention

[0003] The present invention aims to solve the problem that the traditional ship shaft frequency electric field signal feature extraction method in the prior art can no longer solve the feature extraction problem of complex shaft frequency signals under low signal-to-noise ratio conditions.

[0004] The present invention adopts the following scheme: a method for extracting ship shaft frequency electric field characteristics based on eigenmode function energy permutation entropy, comprising the following steps:

[0005] Step 1: Decomposing the original axial frequency electric field signals of the multiple target objects collected by the signal acquisition device into intrinsic mode functions (IMFs) by empirical mode decomposition (EMD); obtaining multiple intrinsic mode function IMF components;

[0006] Step 2: Calculate the instantaneous energy and average energy of each IMF component;

[0007] Step 3: Sort the IMF components according to their average energy values;

[0008] Step 4: Select the IMF component with the largest average energy from the sorted IMF components;

[0009] Step 5: Calculate the permutation entropy (PE) value of the selected maximum average energy IMF component:

[0010] Step 6: Extract the PE value as a signal feature and distinguish the target object.

[0011] Furthermore, the specific steps of step one are as follows:

[0012] (1) Input the original time series signal and use the cubic spline interpolation method to solve the local maximum and minimum values of the signal to form the upper and lower envelopes;

[0013] (2) Calculate the average value of each extreme point in the sequence and subtract the average value from the original signal to obtain h(t); (3) Determine whether h(t) meets the IMF condition. If not, return to step (2) based on h(t) and continue screening until the signal after k decompositions meets the IMF condition, and obtain the first IMF component of the original signal;

[0014] (4) Subtract the first IMF component from the original signal to obtain the residual signal. Repeat step (2) to decompose until the residual signal is a monotonic function that cannot be decomposed. The sum of all IMF components and the residual component is the original signal.

[0015] Furthermore, the target object is a ship, underwater vehicle or other motorized equipment.

[0016] Furthermore, the specific method of step 2 is as follows: since the PE values of the same-order components of different ships are relatively small, the IMF is reordered according to the energy sequence; assuming that the j-th IMF component after the original electric field signal is decomposed has N sampling points, where the instantaneous amplitude of the n-th sampling point is c j (n), then its instantaneous energy is

[0017]

[0018] So the average energy of the jth IMF component is

[0019]

[0020] Furthermore, the maximum average energy in step 4 is defined as assuming that the original time series signal is divided into j IMF components. Defined as:

[0021]

[0022] Where, Represents the average energy of each IMF component after EMD decomposition

[0023] Furthermore, in step 5, the specific steps for calculating the PE value are as follows:

[0024] Reconstruct the phase space of the IMF component with the highest average energy to obtain the reconstruction matrix;

[0025] Sort each row vector of the reconstructed matrix in ascending order to obtain a symbol sequence;

[0026] Count the frequencies of occurrence of each permutation in the symbol sequence to obtain a probability distribution;

[0027] Calculate the PE value according to the probability distribution. The calculation formula of the PE value is:

[0028]

[0029] Furthermore, in step five, the permutation entropy calculation includes the following parameter settings: the time delay τ is 0.5-2.5, and the embedding dimension m is 3-5.

[0030] Furthermore, the signal acquisition device includes:

[0031] Multiple groups of Ag / AgCl electrode units constitute an underwater electric field three-component test system. The electrode spacing is 1 / 2 to 1 / 3 of the ship's width, and the electrode directions are at angles of 0° and 90° to the keel plane, forming a three-component electric field test system.

[0032] In general, the above technical solutions conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0033] (1) The ship shaft frequency electric field feature extraction method based on the eigenmode function energy permutation entropy of the present invention, compared with the PE value and MPE value of the original target signal, the PE value of the maximum average energy IMF as a characteristic parameter can fully reflect the complexity of the target characteristic signal and has better resolution; quantify the complexity and randomness of the signal, thereby effectively distinguishing different ship target signals and solving the problem of different ship signal identification.

[0034] (2) The ship shaft frequency electric field feature extraction method based on the energy permutation entropy of the eigenmode function of the present invention adopts the empirical mode decomposition (EMD) algorithm, which can adaptively decompose the original shaft frequency electric field signal into multiple eigenmode functions (IMFs), effectively extract the local feature information of the signal, retain the original information of the signal to the greatest extent, and provide a basis for subsequent feature extraction.

[0035] (3) The ship shaft frequency electric field feature extraction method based on the eigenmode function energy permutation entropy of the present invention can quickly screen out the IMF components containing the main feature information through energy sorting and PE calculation, improve the calculation efficiency, and realize the rapid identification and classification of the target signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a schematic diagram of the overall process of a preferred embodiment of the invention;

[0037] Figure 2Different ship shaft frequency electric field time domain signals of the preferred embodiment of the invention;

[0038] Figure 3 This is a decomposition diagram of the EMD signals of different ships according to a preferred embodiment of the invention, wherein Figure 3 (a) is the EMD decomposition diagram of the shaft frequency electric field signal of ship A; Figure 3 (b) is the decomposition diagram of the shaft frequency electric field signal EMD of ship B; Figure 3 (c) is the EMD decomposition diagram of the shaft frequency electric field signal of ship C;

[0039] Figure 4 PE value diagram of different ship electric field signal components according to a preferred embodiment of the invention;

[0040] Figure 5 A PE value diagram of IMF sorted by energy according to a preferred embodiment of the invention;

[0041] Figure 6 MPE values of different ships of the preferred embodiment of the invention;

[0042] Figure 7 Comparison of the original signal PE value and the maximum average energy IMF component PE value of the preferred embodiment of the invention;

[0043] Figure 8 is the multi-scale permutation entropy (MPE) of different ship electric field signals of the preferred embodiment of the invention, wherein Figure 8 (a) is the multi-scale permutation entropy when the time scale is 42, Figure 8 (b) is the multi-scale permutation entropy when the time scale is 43;

[0044] Figure 9 A location for experimental testing of preferred embodiments of the invention;

[0045] Figure 10 The shaft frequency electric field signals of different ships measured in the preferred embodiment of the invention;

[0046] Figure 11 The PE value of the maximum average energy IMF of different ships measured in the preferred embodiment of the invention; DETAILED DESCRIPTION

[0047] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only intended to illustrate the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0048] Please refer to Figure 1The present invention relates to a method for extracting ship shaft frequency electric field characteristics based on eigenmode function energy permutation entropy, comprising the following steps:

[0049] Step 1: Decompose the original axial frequency electric field signals of the multiple target objects collected by the signal acquisition device into intrinsic mode functions (IMFs) through empirical mode decomposition (EMD); obtain multiple intrinsic mode function IMF components. The specific steps of step 1 are as follows:

[0050] (1) Input the original time series signal and use the cubic spline interpolation method to solve the local maximum and minimum values of the signal to form the upper and lower envelopes;

[0051] (2) Calculate the average value of each extreme point in the sequence and subtract the average value from the original signal to obtain h(t);

[0052] (3) Determine whether h(t) satisfies the IMF condition. If not, return to step (2) based on h(t) and continue screening until the signal after k decompositions satisfies the IMF condition, and the first IMF component of the original signal is obtained.

[0053] (4) Subtract the first IMF component from the original signal to obtain the residual signal. Repeat step (2) to decompose until the residual signal is a monotonic function that cannot be decomposed. The sum of all IMF components and the residual component is the original signal.

[0054] Step 2: Calculate the instantaneous energy and average energy of each IMF component; the specific method of step 2 is: since the PE values of the same-order components of different ships are slightly different, the IMF is reordered according to the energy sequence; assuming that the j-th IMF component after the original electric field signal is decomposed has N sampling points, where the instantaneous amplitude of the n-th sampling point is c j (n), then its instantaneous energy is

[0055]

[0056] So the average energy of the jth IMF component is

[0057]

[0058] Step 3: Sort the IMF components according to their average energy values;

[0059] Step 4: Select the IMF component with the maximum average energy from the sorted IMF components; the maximum average energy is defined as the highest average energy level if the original time series signal is divided into j IMF components. Defined as:

[0060]

[0061] Where, Represents the average energy of each IMF component after EMD decomposition;

[0062] Step 5: Calculate the permutation entropy (PE) value of the selected maximum average energy IMF component: The specific steps for calculating the permutation entropy value are as follows:

[0063] Reconstruct the phase space of the IMF component with the highest average energy to obtain the reconstruction matrix;

[0064] Sort each row vector of the reconstructed matrix in ascending order to obtain a symbol sequence;

[0065] Count the frequencies of occurrence of each permutation in the symbol sequence to obtain a probability distribution;

[0066] Calculate the PE value according to the probability distribution. The calculation formula of the PE value is:

[0067]

[0068] The permutation entropy calculation includes the following parameter settings: the delay τ is 0.5-2.5, and the embedding dimension m is 3-5.

[0069] Step 6: Extract the PE value as a signal feature and distinguish the target object.

[0070] In some preferred embodiments, the signal acquisition device includes:

[0071] Multiple groups of Ag / AgCl electrode units constitute an underwater electric field three-component test system. The electrode spacing is 1 / 2 to 1 / 3 of the ship's width, and the electrode directions are at angles of 0° and 90° to the keel plane, forming a three-component electric field test system.

[0072] In some other possible embodiments, the signal acquisition device also includes an adaptive adjustment mechanism: a built-in water depth sensor and a salinity sensor, which dynamically adjusts the electrode array spacing through a feedback control module, and the adjustment formula is d=d0·[1+α(S-S0))], where d0 is the reference spacing, S is the real-time salinity, S0 is the reference salinity, and α is a proportional coefficient used to adjust the degree of influence of salinity changes on the electrode spacing.

[0073] In some preferred embodiments, the signal acquisition device also includes a signal preprocessing step, specifically including: band-pass filtering of the original electric field signal, and the passband range is set. The magnetic fixing base includes a temperature compensation module. When the surface temperature of the hull exceeds 40°C, the semiconductor refrigeration plate is automatically started to maintain the electrode working temperature at 25°C±3°C at 0.1Hz-10Hz; an adaptive wavelet threshold method is used to eliminate power frequency interference and marine environmental noise; the signal is smoothed by a sliding average method, and the window length is set to 1 / 10 period of the sampling frequency.

[0074] In some preferred embodiments, a permanent magnet array is integrated at the bottom of the signal acquisition device with a magnetic flux density ≥0.35T, and a flexible conductive rubber pad (Shore hardness 35±5) is configured to achieve non-destructive adsorption and electromagnetic shielding on the hull surface. The magnetic fixing seat includes a temperature compensation module. When the hull surface temperature exceeds 40°C, cooling is automatically started to maintain the electrode operating temperature at 25°C±3°C.

[0075] In order to verify the effectiveness of the algorithm, three types of ships are simulated using three different connection methods of equivalent electric field dipole sources. The experiment uses the same frequency and amplitude signal as input and measures three types of electric field time domain signals as follows: Figure 2 shown.

[0076] Figure 3 The EMD algorithm is used to sequentially decompose the shaft-frequency electric field signals of the three ships into IMFs, from high to low frequency. Analysis of the figure shows that the amplitudes of the IMF1, IMF2, and IMF3 components of Ship A are significantly higher than those of the other components, indicating that they contain the main energy components of the signal characteristics. Secondly, the main energy of Ship B is concentrated in IMF2 and IMF3. The signal energy of Ship C is concentrated in IMF1, IMF2, and IMF3 components.

[0077] To further extract the signal characteristics of different components, the PE value of each IMF decomposition is calculated. Two parameters need to be set when calculating PE: time delay τ and embedding dimension m. Considering the need to retain as many mutation anomalies and characteristic parameters of the actual signal as possible, the time delay τ is set to 1 and the embedding dimension m is set to 4. The PE value calculation results of different ships are shown in the figure below. Figure 4 The X-axis represents the 1st to 8th IMF components after the electric field signal is decomposed, and the Y-axis represents the PE value of each component.

[0078] As can be seen from the figure, the PE values of the shaft-frequency electric field signal IMF components of the three models decrease with increasing order, indicating that as the order of the IMF increases, the complexity of the component signal decreases. Furthermore, the PE values of the shaft-frequency electric field signal IMF components of the same ship vary significantly. However, compared to the electric field signal of the same ship, the PE values of the same-order components of different ships do not vary significantly.

[0079] Since the PE values of the same-order components of different ships are relatively small, the IMFs are reordered according to the energy sequence. Assume that the j-th IMF component after the original electric field signal is decomposed has N sampling points, where the instantaneous amplitude of the n-th sampling point is c j (n), then its instantaneous energy is

[0080]

[0081] So the average energy of the jth IMF component is

[0082]

[0083] According to equations (11) and (12), the energy value of each IMF component is calculated as follows: Figure 5 As shown in the figure, the X-axis represents the eight IMF components sorted by energy from high to low, and the Y-axis represents the PE value of each component.

[0084] The results show that there is no clear pattern between the energy values of the IMF components and their PE values. There are significant differences in PE values between different IMF components for the same ship, but for different ships, the PE values after the fifth energy level are not significantly different. Based on the above analysis, the PE values of the first three energy levels vary significantly for different ships, and for time series signals, the larger the energy value, the more signal characteristic main components it contains.

[0085] By analyzing the PE values of the IMFs at various levels of the three different electric field signals, the differences are primarily concentrated in the first three energy levels. The IMF component of the first energy level contains the main signal characteristics and exhibits significant differences in PE values. While the PE values of the second and third energy levels differ significantly, the signal characteristics are not as comprehensive as those of the first energy level. Therefore, the present invention uses the IMF component of the first energy level to determine its characteristics.

[0086] Assuming that the original time series signal is divided into j IMF components, the highest average energy level is defined as:

[0087]

[0088] Where, Represents the average energy of each IMF component after EMD decomposition. The highest average energy and PE value of each IMF component of ship A, ship B and ship C are calculated according to equations (11)-(13) and are shown in Table 1.

[0089] Table 1 Component PE values of three different types of ships

[0090]

[0091] According to the analysis in Table 1, the maximum average energy levels of the decomposed signals of different ships are different, and the average energy difference is greater than 0.02, and the PE difference is greater than 0.3.

[0092] Since the complexity of time series signals of signals of different types of ships in the same sea area is completely different, the present invention uses multiscale permutation entropy (MPE) to describe the complexity of time series at different scales, such as Figure 6 As shown. The embedding dimension is 4, the time delay is 1, the X-axis represents the time scale, and the Y-axis represents the entropy value corresponding to each time scale.

[0093] As can be seen from the figure, as the time scale increases, ships B and C fluctuate within a certain range with insignificant amplitudes, while the amplitude of ship A fluctuates significantly. The three signals differ somewhat at different time scales, with the greatest differences occurring at time scales of 42 and 43.

[0094] Table 2 Entropy values at different time scales

[0095]

[0096] Table 2 analyzes the entropy values of the three ship signals at time scales 42 and 43. Within both time scales, the entropy values of ship A differ significantly from those of the other two ships, while the difference between ships B and C is very small. Overall, the differences are not significant, and further analysis of separability is needed.

[0097] In order to verify the universality of the method of the present invention, 100 samples of different types of signals were taken as test data to compare the effects of three different feature extraction methods. The PE value of the original signal and the PE value of the maximum average energy IMF were extracted as feature parameters, respectively. Figure 7 shown.

[0098] Depend on Figure 7 (a) Analysis shows that there is a significant difference between ship A and the other two ships, about 0.2, but the difference between ships B and C is less than 0.05, making it difficult to distinguish their characteristic parameters. Figure 7 In (b), the entropy values of the IMF components of the maximum average energy of the same ship are basically maintained at the same level, and the PE values between different ships differ by more than 0.2, indicating that the characteristic parameters of the shaft frequency electric field of different ships can be effectively identified.

[0099] For the MPE values of different ships at different time scales, the above analysis shows that the entropy value difference is the largest when the time scale is 42 and 43, so the time scales are selected as 42 and 43 respectively. The experimental results are as follows: Figure 8 shown.

[0100] Depend on Figure 8(a) Analysis shows that when the time scale is 42, the entropy difference between ships A and B is about 0.13, while the entropy difference between ships A and C is less than 0.1, which cannot be effectively distinguished; similarly Figure 8 (b) The time scale is 43, and the difference between ships B and C is about 0.05, which makes it difficult to accurately distinguish their characteristic differences.

[0101] Actual ship data verification

[0102] In order to further verify the effectiveness of the method of the present invention, a sea trial was conducted near Zhifu Island, Yantai (water depth of about 17m, seawater conductivity of about 3.5S / m, longitude 121.43°, latitude 37.58°). Figure 9 In the experiment, Ag / AgCl electrodes were used, with an electrode distance of about 1m and a sampling frequency of 100Hz. The measurement system was arranged near the channel. The basic parameters of three different types of ships are shown in Table 3. The speed was estimated from the measurement value of the shore-based distance meter, and its time domain shaft frequency signal is shown in Figure 10 As shown, its entropy value is Figure 11 shown.

[0103] Table 3 Information of tested vessels

[0104]

[0105] The algorithm proposed in this paper calculates entropy using actual ship electric field signal data. The graph shows significant differences in PE values for different ships. The difference between ships A and B is approximately 0.5, while the difference between ships A and C is approximately 0.3, demonstrating that the proposed method can effectively distinguish target signals.

[0106] Table 4 Characteristic parameters of electric field signals of different ships

[0107]

[0108] According to the data analysis in Table 4, the variance of the entropy values for each ship is very small across the three feature extraction methods: PE value, PE value of maximum mean energy (IMF), and MPE value, indicating that the fluctuation range of entropy values for different ships is relatively stable. The PE value feature parameter for the maximum mean energy (IMF) fluctuates within different ranges, with the range difference between ships A and B being approximately 0.53, and the range difference between ships A and C being approximately 0.22, demonstrating good discrimination. Compared to the PE value feature extraction method using the maximum mean energy (IMF), when using only the PE value as a feature parameter, the maximum difference between ships A and C is only approximately 0.005, making it difficult to distinguish the signal characteristics of different ships. When using the MPE value as a feature parameter, the fluctuation ranges of ships B and C partially overlap at a time scale of 34, making them indistinguishable. In summary, the PE value of the maximum mean energy (IMF) better distinguishes the three signals than the other two feature extraction methods.

[0109] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for extracting ship shaft frequency electric field features based on eigenmode function energy permutation entropy, characterized in that: The steps include: Step 1: decomposing the original axial frequency electric field signals of the target objects collected by the signal acquisition device into intrinsic mode functions by empirical mode decomposition; obtaining multiple intrinsic mode function IMF components; Step 2: Calculate the instantaneous energy and average energy of each IMF component; Step 3: Sort the IMF components according to their average energy values; Step 4: Select the IMF component with the largest average energy from the sorted IMF components; Step 5: Calculate the permutation entropy (PE) value of the selected maximum average energy IMF component: Step 6: Extract the PE value as a signal feature and distinguish the target object.

2. The method for extracting ship shaft frequency electric field characteristics based on eigenmode function energy permutation entropy according to claim 1 is characterized by: The specific steps of step one are as follows: (1) Input the original time series signal and use the cubic spline interpolation method to solve the local maximum and minimum values of the signal to form the upper and lower envelopes; (2) Calculate the average value of each extreme point in the sequence and subtract the average value from the original signal to obtain h(t); (3) Determine whether h(t) satisfies the IMF condition. If not, return to step (2) based on h(t) and continue screening until the signal after k decompositions satisfies the IMF condition, and the first IMF component of the original signal is obtained. (4) Subtract the first IMF component from the original signal to obtain the residual signal. Repeat step (2) to decompose until the residual signal is a monotonic function that cannot be decomposed. The sum of all IMF components and the residual component is the original signal.

3. The method for extracting ship shaft frequency electric field characteristics based on eigenmode function energy permutation entropy according to claim 1 is characterized by: The target object is a ship, an underwater vehicle or other motorized equipment.

4. The method for extracting ship shaft frequency electric field characteristics based on eigenmode function energy permutation entropy according to claim 1 is characterized by: The specific method of step 2 is as follows: since the PE values of the same-order components of different ships are relatively small, the IMF is reordered according to the energy sequence; assuming that the j-th IMF component after the original electric field signal is decomposed has N sampling points, where the instantaneous amplitude of the n-th sampling point is c j (n), then its instantaneous energy is So the average energy of the jth IMF component is 5. The method for extracting ship shaft frequency electric field characteristics based on eigenmode function energy permutation entropy according to claim 1 is characterized in that: The maximum average energy in step 4 is defined as assuming that the original time series signal is divided into j IMF components. Defined as: Where, Represents the average energy of each IMF component after EMD decomposition.

6. The method for extracting ship shaft frequency electric field characteristics based on eigenmode function energy permutation entropy according to claim 1 is characterized by: In step 5, the specific steps for calculating the permutation entropy value are as follows: Reconstruct the phase space of the IMF component with the highest average energy to obtain the reconstruction matrix; Sort each row vector of the reconstructed matrix in ascending order to obtain a symbol sequence; Count the frequencies of occurrence of each permutation in the symbol sequence to obtain a probability distribution; Calculate the PE value according to the probability distribution. The calculation formula of the PE value is:

7. The method for extracting ship shaft frequency electric field characteristics based on eigenmode function energy permutation entropy according to claim 1 is characterized by: In step 5, the permutation entropy calculation includes the following parameter settings: the time delay τ is 0.5-2.5, and the embedding dimension m is 3-5.

8. The method for extracting ship shaft frequency electric field characteristics based on eigenmode function energy permutation entropy according to any one of claims 1 to 7, characterized in that: The signal acquisition device comprises: Multiple groups of Ag / AgCl electrode units constitute an underwater electric field three-component test system. The electrode spacing is 1 / 2 to 1 / 3 of the ship's width, and the electrode directions are at angles of 0° and 90° to the keel plane, forming a three-component electric field test system.

Images