Micro-motion target feature extraction method based on multi-domain parameter estimation and phase compensation

Abstract

The invention provides a micro-motion target feature extraction method based on multi-domain parameter estimation and phase compensation. The micro-motion target feature extraction method comprises the following steps: analyzing periodic peak features existing in a time domain and a frequency domain of a rotor target; extracting a plurality of peak values of a time domain and a frequency domain through a sliding window peak value search algorithm, and realizing estimation of a time domain peak value frequency and a frequency domain spectrum interval by utilizing the extracted peak values; through comparison of the time domain peak frequency and the frequency domain spectrum interval number, the odd-even number of target blades of the rotor wing is determined, and rough estimation of the blade number and the rotating speed parameter of the target is achieved; a phase compensation factor is constructed by using the estimated blade number of the target and the rotating speed parameter to perform phase compensation on the target echo, and the target blade number is determined by analyzing the compensated signal time frequency result, so that the accurate estimation of the rotor target blade number and the rotating speed parameter is realized, and the accurate characteristics of the rotor target are obtained. The method can effectively and quickly extract rotor target features, has more stable performance, and can provide a large amount of data support for rotor target identification.

Description

Technical Field

[0001] The present invention relates to the technical field of radar signals, and in particular, to a method for extracting micro-motion target features with multi-domain parameter estimation and phase compensation. Background Art

[0002] There are mainly four types of methods for extracting the micro-motion features of multi-component periodic micro-Doppler signals: (1) Methods based on signal decomposition, that is, decomposing a signal containing multiple signal components into multiple single-component signals. However, there are too many micro-Doppler components, and there is interference between components, making it difficult to effectively separate the signal components and then extract the micro-motion features; (2) Methods based on the transform domain, by seeking various domain transformation methods to remove redundant features and more efficiently extract signal features. However, for rotor targets, these methods change the feature dimensions, and the physical meanings of the extracted features are not well explained; (3) Methods based on the image domain, usually using the short-time Fourier transform to transform the signal into the time-frequency domain, and then using the transform to convert the curve detection problem in the image space into a peak detection problem in the parameter space to extract the micro-motion features of the target. However, for rotor targets, due to the existence of the scintillation phenomenon, these time-frequency transformation methods cannot effectively extract features; (4) Methods based on sparse reconstruction, which use orthogonal matching pursuit algorithms to invert the micro-Doppler components. However, in the scattering point model of rotor targets, the number of scattering points is too large to use this method for inversion.

[0003] There are a large number of signal components in rotor targets, and there is interference between signals. Therefore, a scintillation phenomenon will occur in the time-frequency result after signal superposition, and the main energy of the signal is concentrated on the scintillation, resulting in the ineffectiveness of the method of using the signal sine curve for feature extraction in low signal-to-noise ratio or clutter situations. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for extracting micro-motion target features with multi-domain parameter estimation and phase compensation in view of the above-mentioned deficiencies of the prior art.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: The present invention provides a method for extracting micro-motion target features with multi-domain parameter estimation and phase compensation, including the following steps: S1. Analyze the periodic peak features existing in the time domain and frequency domain of the rotor target; S2. Extract multiple peaks in the time domain and frequency domain through a sliding window peak search algorithm, and use the extracted peaks to estimate the time domain peak frequency and the frequency domain spectrum interval; S3. Determine the odd or even number of rotor target blades by comparing the time domain peak frequency and the number of frequency domain spectrum intervals, and roughly estimate the number of blades and the rotational speed parameters of the target; S4. Construct a phase compensation factor using the estimated number of blades and rotational speed parameters of the target to perform phase compensation on the target echo. Determine the number of target blades by analyzing the time-frequency result of the compensated signal, thereby achieving accurate estimation of the number of blades and rotational speed parameters of the rotor target and obtaining relatively accurate characteristics of the rotor target.

[0006] Further, the specific content of S1 is as follows: For the th blade, there will be two peaks in one period. Let the moments when the peaks appear be and respectively. Then the expressions for these two moments are: (1); (2); Among them, is the total number of rotor blades; is the th blade; is the initial phase of the first blade; is the blade rotation rate of the rotor target, is the angular velocity of rotation.

[0007] Further, when calculating the number of peaks in the echo of a rotor-like target, for a rotor target with the number of blades being , there will be a moment when peaks exist simultaneously on different blades. Therefore, the time-domain peak frequency is not necessarily . Then, the number of blades is analyzed by dividing them into odd and even numbers of blades: When the number of blades is odd, let . At this time, the moment when the th blade appears a peak is: (3); (4); Verify whether there is a moment when the th blade and the th blade appear peaks simultaneously in one period. If it exists, it satisfies , that is: (5); The left-side expression of formula (5) is even, and the right-side expression is odd, that is, does not hold. Then, when the rotor target is an odd-blade target, there is no moment when different blades appear peaks simultaneously. At this time, the number of peaks appearing in one period is , and the time-domain peak frequency is , which is twice the spectral interval; When the number of blades is even, let . At this time, the The moment when a peak appears in a blade is: (6); (7); Verify whether there is a moment when different blades simultaneously appear at peaks within one cycle. If so, it satisfies , that is: (8); Equation (8) has a condition for the equation to hold, that is , at this time ; Since is half of the number of blades, then the th blade and the th blade have the same corresponding moments when time-domain peaks appear. At this time, the number of peaks appearing within one cycle is , and the time-domain peak frequency is , which is the same as the spectral interval; From equations (5) and (8), the time interval between two peaks in the time domain satisfies: (9).

[0008] Furthermore, the specific content of S2 is: The time-domain sliding window peak search algorithm is: When the sampling frequency is , and the echo length is , the echo of the rotor target is expressed as: (10); S201. Use the window length threshold search and estimation algorithm to estimate the planned window length and the peak threshold ; S202. Perform sliding window calculation to extend the echo signal to obtain the extended signal ; (11); At this time, perform peak search on the extended signal. Specifically: Take the first moment of the signal, and the signal amplitude corresponding to this moment is ; Take out the sliding window signal , and the formula is: (12); Obtain the maximum value of the sliding window signal. If the first moment , and the signal echo amplitude simultaneously satisfies the conditions and , it is considered that the first moment has a peak, and the first moment is recorded in the peak set , the first peak moment ; if the condition is not met, the first moment is not included in the peak set; S203. Perform the S2 operation on the signal moments in sequence, and finally obtain the peak set: (13); is the number of moments satisfying the condition, that is, the number of judged peaks, which completes the time-domain peak sliding window search.

[0009] Furthermore, the S2 further includes: The frequency-domain sliding window peak search algorithm is: Perform moving target indication (MTI) operation on the rotor target echo to filter out the frequency components near zero frequency. After the echo with a length of passes through the MTI, perform a fast Fourier transform to obtain the frequency domain: (14); S21. Use the window length threshold search estimation algorithm to estimate the window length and the peak threshold ; S22. Perform sliding window calculation to extend the echo signal to obtain the extended signal ; (15); At this time, perform peak search on the extended signal. The steps are as follows: Take the frequency of the first frequency point , and the signal amplitude corresponding to this moment is . Take out the sliding window signal as: (16); Obtain the maximum value of the sliding window signal. If the frequency point , and the signal echo amplitude simultaneously satisfies the conditions and , it is considered that the frequency point has a peak, and the frequency point is recorded in the peak set , the first peak frequency point . If the condition is not met, the frequency point is not included in the peak set; S23. Perform the operation of S22 on the signal moments in sequence, and finally obtain a peak set: (17); The number of moments that meet the conditions, that is, the number of judged peaks, is searched, which means the frequency-domain sliding window peak search is completed.

[0010] Further, the specific content of S3 is as follows: Through the sliding window peak search, a time-domain peak set and a frequency-domain peak set are obtained. The time-domain peak number and the frequency-domain spectrum interval are obtained by using the peak set. Since there is noise and clutter in the measured data, calculate the time difference between two adjacent time-domain peaks in the time-domain peak set , and the number of time differences is 1 less than the peak number. Then take the reciprocal of the time difference to obtain the time-domain peak frequency set : (18); Set a certain acceptable estimation error range, and finally the time-domain peak number is the average of the acceptable results to obtain the estimated time-domain peak frequency ; For the calculation of the frequency-domain peak number, by calculating the frequency difference between two adjacent frequency-domain peaks in the frequency-domain peak set , and the number of frequency differences is 1 less than the peak number. Then the time difference is the frequency-domain spectrum interval, and the frequency-domain spectrum interval set is obtained: (19); Set a certain acceptable estimation error range, and finally the frequency-domain spectrum interval is the average of the acceptable results to obtain the estimated spectrum interval ; By comparing the estimated time-domain peak frequency and the estimated spectrum interval , the parity of the target blade number can be determined. If , the target is an odd-blade; if , the target is an even-blade, that is, the target rotation speed and the blade number can be roughly estimated.

[0011] Further, the specific content of S4 is as follows: For the rotor target, the echo is generated by the superposition of the echoes of multiple blades with different initial phases. That is, each blade echo is divided into two parts , : (20); Perform short-time Fourier transform on the signal, convert the product of time-domain signals into frequency-domain convolution, and the transformation result is: (21); Perform phase compensation on the signal part, and construct the phase compensation factor : (22); wherein, is the variable value in the compensation factor, and ; (23); The compensated rotor target echo signal is: (24); Compare the echo after phase compensation with the echo without compensation, and also divide the echo after phase compensation into two parts: , : (25).

[0012] Furthermore, when the number of blades is odd, let , and the compensated echo at this time is: (26); It can be seen from equation (26) that only when , , the phase compensation of the first blade is realized. For other blades, compared with before phase compensation and are exactly the same. At this time, the changed is manifested in the time-frequency result as: The sine amplitude changes from to ; The phase of the sine curve changes, from changing to ; After phase compensation the time-frequency result is convolved with the time-frequency result, and the time-frequency result of each blade obtained has an obvious difference from that before phase compensation; The moment when each blade flashes does not change, and at the same time, the amplitude of the sine curve is affected. The amplitudes of the two sine curves corresponding to the th blade are: (27).

[0013] The flashing of each blade is periodic. For odd-numbered blades, after phase compensation, only one blade can be fully phase-compensated. At this time, the periodic characteristics in the micro-motion flashing frequency band of the rotor component can be utilized to search for the number of flashing blades by determining the number of flashes between the flashing frequency bands symmetric about zero frequency after phase compensation.

[0014] Furthermore, when the number of blades is even, let , and the compensated echo at this time is: (28); It can be seen from Equation (28) that when , at this time , the phase compensation of the first blade is achieved; when the number of target blades of the rotor is even, at each moment when there is flashing, there will be two blades flashing simultaneously. At this time, one blade has positive-frequency flashing and one blade has negative-frequency flashing. The two blades , satisfy ; When , (29); When , (30).

[0015] Furthermore, after phase compensation the time-frequency result is convolved with the time-frequency result, and the time-frequency result of each blade obtained has an obvious difference from that before phase compensation. The moment when each blade flashes has changed. The amplitudes of the two sine curves corresponding to the th blade are: (31); At this time, except for the and blades, the amplitudes and initial phases of the two curves in the time-frequency results of other blades are different, and the flashing frequency band is not symmetric about the zero frequency of 0; For the blade, the amplitudes of the two corresponding sine curves in the time-frequency result are both , and the amplitudes of the two sine curves corresponding to the blade are: In the time-frequency result, the flashing frequency band is located between the two sine curves, and the amplitudes of the two curves are and , and the two sine curves reach the maximum and minimum values simultaneously; While is at, at this time When superimposing the time-frequency results, the amplitudes of the two sine curves are as follows: (33); The flickering of the two blades is superimposed in the time-frequency result, and the two curves reach the extreme values simultaneously. When one sine curve reaches the maximum value, the other sine curve reaches the minimum value. At this time, the bandwidth of the flickering frequency band is A , and the range is , and the ratio of the bandwidth of the flickering frequency band on the positive half-axis to the bandwidth of the flickering frequency band on the negative half-axis is approximately .

[0016] The beneficial effects of the present invention are as follows: First, the time-domain and frequency-domain peak characteristics of the rotor target are analyzed theoretically, and based on this theoretical basis, a multi-peak search algorithm is designed to estimate the time-domain peak frequency and the number of spectral intervals in the frequency domain, so as to determine the odd and even numbers of the rotor target blades, and then roughly estimate the parameters such as the number of blades and the rotational speed of the rotor target. On this basis, using the estimated blade rotational speed and blade number information, a phase compensation factor is constructed, and the number of rotor target blades is estimated through the compensated signal echo. Description of the Drawings

[0017] Figure 1 is a flow chart of a method for extracting the characteristics of a micro-moving target with multi-domain parameter estimation and phase compensation; Figure 2 is a micro-motion model diagram of a helicopter rotor; Figure 3 is a disassembled diagram of the time-frequency result of odd-numbered blades after phase compensation; Figure 4 is a disassembled diagram of the time-frequency result of even-numbered blades after phase compensation; Figure 5 (a) is the time-domain peak search result of the helicopter Mi-17; Figure 5 (b) is the frequency-domain peak search result of the helicopter Mi-17; Figure 6 is the helicopter feature estimation result when the sampling duration is 0.2 s. Detailed Embodiment

[0018] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0019] Please refer to Figure 1 , a method for extracting the characteristics of a micro-moving target with multi-domain parameter estimation and phase compensation, comprising the following steps: S1. Analyze the periodic peak characteristics existing in the time domain and frequency domain of the rotor target; S2. Extract multiple peaks in the time domain and frequency domain through a sliding window peak search algorithm, and use the extracted peaks to estimate the time domain peak frequency and the frequency domain spectrum interval; S3. Determine the odd or even number of the target rotor blades by comparing the time domain peak frequency and the number of frequency domain spectrum intervals, and roughly estimate the number of blades and the rotational speed parameters of the target; S4. Use the estimated number of blades and rotational speed parameters of the target to construct a phase compensation factor to perform phase compensation on the target echo. Determine the number of target blades by analyzing the time-frequency result of the compensated signal, and then achieve the accurate estimation of the number of rotor blades and rotational speed parameters of the target, and obtain the characteristics of the relatively accurate rotor target.

[0020] The relative position relationship between the rotor target and the radar is as Figure 2 shown, After the rotor echo signal under the scattering point model is matched filtered and translational compensated, the baseband echo is expressed as: (101); where, is the imaginary unit, is the total number of rotor blades, is the blade length, is the initial phase of the first blade. The micro-Doppler frequency of the rotor target is time-varying. For the th blade, its micro-Doppler frequency expression is: (102); From equation (102), it can be obtained that the micro-Doppler frequency of the target echo signal is modulated by the sine function , that is, the micro-Doppler frequency is time-varying and shows non-linear variation. The maximum micro-Doppler frequency of the target is jointly determined by the length, rotational speed, pitch angle of the rotor and the radar carrier wavelength; Perform Fourier transform on equation (102) to obtain the frequency domain characteristics of JEM expressing the ideal rotor echo: (103); It can be seen from equation (3) that the echo in the frequency domain is the sum of a series of impulse functions, that is, centered on , and consists of a series of line spectra with a spectrum interval of , is the number of effective sidebands, and its number of sidebands is the integer part of the formula . The spectrum interval is determined by the number of blades and the rotational speed . According to theoretical analysis, the spectrum interval is ; the spectrum line amplitude is determined by the parameter and the Bessel function.

[0021] In the time domain, for a blade, there will be two peaks in one period. Therefore, the peak frequency in 1 s is .

[0022] The echo includes the time-domain information of the echoes of each blade. Therefore, the number of peaks appearing in the time domain within 1 s is related to the number of blades and the rotational speed.

[0023] The specific S1 is as follows: For the th blade, there will be two peaks in one period. Let the moments when the peaks appear be and respectively. Then the expressions for the two moments are: (1); (2); where is the total number of rotor blades; is the th blade; is the initial phase of the first blade; is the blade rotation rate of the rotor target, is the angular rotation velocity.

[0024] When calculating the number of echo peaks of a rotor-like target, for a rotor target with blades, there will be a moment when peaks exist simultaneously on different blades. Therefore, the time-domain peak frequency is not necessarily . Then the blades are analyzed by dividing them into odd and even numbers: When the number of blades is odd, let . At this time, the moment when the th blade appears a peak is: (3); (4); Verify whether there is a moment when the th blade and the th blade appear peaks simultaneously within one period. If it exists, it satisfies , that is: (5); The left-side expression of formula (5) is an even number, and the right-side expression is an odd number, that is, does not hold. Then when the rotor target is an odd-blade target, there is no moment when peaks appear simultaneously on different blades. At this time, the number of peaks appearing in one period is , and the time-domain peak frequency is , which is twice the spectral interval; When the number of blades is even, let , at this time, the moment when the th blade appears at the peak is: (6); (7); Verify whether there is a moment when different blades appear at the peak simultaneously within one cycle. If so, it satisfies , that is: (8); There is a condition for the equation in formula (8) to hold, that is , at this time ; Since is half of the number of blades, then the th blade and the th blade have the same corresponding moments when they appear at the time-domain peak. At this time, the number of peaks appearing within one cycle is , and the time-domain peak frequency is , which is the same as the spectral interval; From formulas (5) and (8), it can be obtained that the time interval between the two peaks in the time domain satisfies: (9).

[0025] The specific S2 is as follows: The time-domain sliding window peak search algorithm is: When the sampling frequency is , and the echo length is of the rotor target echo is expressed as: (10); S201. Use the window length threshold search and estimation algorithm to estimate the window length and the peak threshold ; S202. Perform sliding window calculation to extend the echo signal to obtain the extended signal ; (11); At this time, perform peak search on the extended signal, specifically: Take the first moment of the signal, and the amplitude of the signal at this corresponding moment is ; Take out the sliding window signal , and the formula is: (12); Find the maximum value of the sliding window signal. If the first moment , the signal echo amplitude simultaneously satisfies the conditions and , then it is considered that the peak appears at the first moment , and the first moment is recorded in the peak set , the first peak moment ; if the conditions are not satisfied, the first moment is not included in the peak set; S203. Perform the S2 operation on the signal moments in sequence, and finally obtain the peak set: (13); is the number of moments satisfying the conditions, that is, the number of judged peaks, and the time-domain peak sliding window search is completed.

[0026] The S2 also includes: The frequency-domain sliding window peak search algorithm is: Perform the moving target indication (MTI) operation on the rotor target echo to filter out the frequency components near zero frequency. After the echo with a length of passes through the MTI, perform the fast Fourier transform to obtain the frequency domain: (14); S21. Use the window length threshold search estimation algorithm to estimate the planned window length and the peak threshold ; S22. Perform the sliding window calculation to extend the echo signal to obtain the extended signal ; (15); At this time, perform peak search on the extended signal, and the steps are: Take the frequency of the first frequency point , and the signal amplitude at the corresponding moment is , take out the sliding window signal as: (16); Obtain the maximum value of the sliding window signal. If the frequency point , the signal echo amplitude simultaneously satisfies the conditions and , then it is considered that the peak appears at the frequency point , and the frequency point is recorded in the peak set , the first peak frequency point , if the conditions are not satisfied, the frequency point is not included in the peak set; S23. Perform the operation of S22 on the signal moments in sequence, and finally obtain a peak set: (17); To search for the number of moments that meet the conditions, that is, the number of judged peaks, namely, complete the frequency-domain sliding window peak search.

[0027] The specific content of S3 is as follows: Through the sliding window peak search, obtain the time-domain peak set and the frequency-domain peak set, and use the peak set to obtain the time-domain peak number and the frequency-domain spectrum interval; due to the existence of noise and clutter in the measured data, calculate the time difference between two adjacent time-domain peaks in the time-domain peak set , the number of time differences is 1 less than the number of peaks, and then take the reciprocal of the time difference to obtain the time-domain peak frequency set : (18); Set a certain acceptance estimation error range, and finally the time-domain peak number is the average of the acceptable results to obtain the estimated time-domain peak frequency ; For the calculation of the frequency-domain peak number, by calculating the frequency difference between two adjacent frequency-domain peaks in the frequency-domain peak set , the number of frequency differences is 1 less than the number of peaks, and then the time difference is the frequency-domain spectrum interval to obtain the frequency-domain spectrum interval set : (19); Set a certain acceptance estimation error range, and finally the frequency-domain spectrum interval is the average of the acceptable results to obtain the estimated spectrum interval ; By comparing the estimated time-domain peak frequency and the estimated spectrum interval , the parity of the target blade number can be determined. If , the target is an odd-numbered blade; if , the target is an even-numbered blade, that is, the target rotational speed and the blade number can be roughly estimated.

[0028] The specific content of S4 is as follows: For the rotor target, the echo is generated by the superposition of the echoes of multiple blades with different initial phases. That is, each blade echo is divided into two parts , : (20); Perform the short-time Fourier transform on the signal, convert the time-domain signal product into a frequency-domain convolution, and the transformation result is: (21); Perform phase compensation on the signal portion, and construct a phase compensation factor : (22); Wherein, is the variable value in the compensation factor, and ; (23); The compensated rotor target echo signal is: (24); Compare the echo after phase compensation with the echo without compensation, and also divide the echo after phase compensation into two parts: , : (25).

[0029] When the number of blades is odd, let , and the echo after compensation at this time is: (26); It can be seen from Equation (26) that only when , , at this time, the phase compensation of the first blade is achieved. For other blades, compared with before phase compensation and are exactly the same. At this time, the changed is manifested in the time-frequency result as: The sine amplitude changes from to ; The phase of the sine curve changes, from changing to ; After phase compensation the time-frequency result is convolved with the time-frequency result, and the time-frequency result of each blade obtained has an obvious difference from that before phase compensation; The moment when each blade flashes does not change, and at the same time, the amplitude of the sine curve is affected. The amplitudes of the two sine curves corresponding to the th blade are: (27).

[0030] The flashing of each blade is periodic. For odd-numbered blades, after phase compensation, only one blade can be fully phase-compensated. At this time, the periodic characteristics in the micro-motion flashing frequency band of the rotor component can be utilized to search for the number of flashing blades by determining the number of flashes between the flashing frequency bands symmetric about the zero frequency after phase compensation. Taking a 5-blade rotor as an example, the time-frequency result of the echo is disassembled as follows Figure 3 as shown.

[0031] When the number of blades is even, let , and the compensated echo at this time is: (28); It can be seen from Equation (28) that by setting , at this time , and the phase compensation of the first blade is achieved at this time; when the number of target blades of the rotor is even, at each moment when flashing occurs, there will be two blades flashing simultaneously. At this time, one blade has positive-frequency flashing and one blade has negative-frequency flashing. The two blades 、 satisfy ; When , (29); When , (30).

[0032] After phase compensation the time-frequency result is convolved with the time-frequency result, and the time-frequency result of each blade obtained has an obvious difference from that before phase compensation. The moment when each blade flashes has changed. The amplitudes of the two sine curves corresponding to the th blade are: (31); At this time, except for the blades of and , the amplitudes and initial phases of the two curves in the time-frequency results of other blades are different, and the flashing frequency bands are not symmetric about the 0 zero frequency; For blade , the amplitudes of the two corresponding sine curves in the time-frequency result are both , The amplitudes of the two sine curves corresponding to the blade are: (32); In the time-frequency result, the flashing frequency band is located between the two sine curves, and the amplitudes of the two curves are and , and the two sine curves reach the maximum and minimum values simultaneously; while at this time and when superimposing with the time-frequency result, the amplitudes of the two sine curves are: (33); The flickering of the two blades is superimposed in the time-frequency result, and the two curves reach the extreme values simultaneously. When one sine curve reaches the maximum value, the other sine curve reaches the minimum value. At this time, the bandwidth of the flickering frequency band is A , and the range is , and the ratio of the bandwidth of the positive half-axis flickering frequency band to the bandwidth of the negative half-axis flickering frequency is approximately .

[0033] The flickering of each blade is periodic. For even-numbered blades, after phase compensation, only one blade can be fully phase-compensated. Similar to odd-numbered blades, at this time, the periodic characteristics of the micro-motion flickering frequency band of the rotor component can be used to search for the ratio of the positive and negative flickering frequency band bandwidths after phase compensation as to determine the number of flickering blades based on the number of flickers between the flickering frequency bands. Taking a 4-blade as an example, the time-frequency result of the echo is disassembled as Figure 4 shown

[0034] Embodiment 1 Select 4 types of helicopters and 4 types of propeller aircraft for simulation. The aircraft model parameters are shown in Table 1. Other parameter settings: the sampling frequency is , and the carrier wavelength is .

[0035] Table 1 Rotor model parameters of helicopter aircraft

[0036] Taking the Mi17 helicopter as an example, the peak search result is as Figure 5 shown

[0037] Table 2 shows the number of target time-domain peaks and the estimated results of the frequency-domain spectral interval under different signal-to-noise ratios at the sampling time .

[0038] Table 2 Estimated results of time-domain peak frequencies under different signal-to-noise ratios (Hz)

[0039] As Figure 6 shown, for different sampling times, the estimated results of all simulation targets are shown in Table 3. It is possible to compare the estimated time-domain peak with the estimated frequency-domain spectral interval to judge the parity of the target blade

[0040] Table 3 Estimated results of spectral interval and time-domain peak number

[0041] As can be seen from Table 4, through the calculation of the peak spectrum interval, the number of blades and the rotational speed of the target can be roughly estimated, which will provide important information for target recognition.

[0042] Table 4 Estimation results of target rotational speed and number of blades

[0043] To estimate the number of blades of a rotor target, a phase compensation factor needs to be constructed. The spectral interval of the rotor target is estimated through a peak search algorithm. On this basis, the signal is transformed into the frequency domain, and the variables in the phase compensation factor are obtained according to the signal frequency domain range. The specific parameters of 8 rotor targets are shown in Table 5. Since the estimation results of the rotational speed and the number of blades obtained by peak search cannot accurately determine the number of blades of the rotor target, there are two sets of variable compensation factors for the constructed compensation factor. For the variables in the compensation factor, iterative matching can be performed.

[0044] Table 5 Rough estimation results of the characteristic parameters of air targets

[0045] After determining the number of blades, the rotational speed of the rotor target can be determined, and according to the micro-Doppler frequency range and the pitch angle, the blade length of the rotor target can be determined to realize the extraction of the rotor target characteristics. The characteristic extraction results and the theoretical results are shown in Table 6. It can be shown from Table 6 that the characteristic extraction method proposed in this paper can effectively extract the characteristics of the rotor target by using the target echo signal and realize the recognition of the rotor target.

[0046] Table 6 Characteristic extraction results of air targets

[0047] The embodiments described above only represent the implementation manners of the present invention. The description is relatively specific and detailed, but it cannot be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention should be based on the appended claims.

Claims

1. A method for extracting micro-motion target features with multi-domain parameter estimation and phase compensation, characterized in that, Including the following steps: S1. Analyze the periodic peak characteristics existing in the time domain and frequency domain of the rotor target; S2. Extract multiple peaks in the time domain and frequency domain through the sliding window peak search algorithm, and use the extracted peaks to estimate the time domain peak frequency and the frequency domain spectrum interval; S3. Determine the odd and even numbers of the rotor target blades by comparing the time domain peak frequency and the number of frequency domain spectrum intervals, and roughly estimate the number of blades and the rotational speed parameters of the target; S4. Construct a phase compensation factor using the estimated number of blades and rotational speed parameters of the target to perform phase compensation on the target echo. Determine the number of target blades by analyzing the time-frequency result of the compensated signal, and then achieve the accurate estimation of the number of rotor target blades and the rotational speed parameters, and obtain relatively accurate characteristics of the rotor target.

2. The feature extraction method for micro-motion target with multi-domain parameter estimation and phase compensation according to claim 1, characterized in that The specific content of S1 is as follows: For the th blade, there will be two peaks in one cycle. Let the moments when the peaks appear be , , then the expressions for the two moments are: （1）； （2）； Among them, is the total number of rotor blades; is the th blade; is the initial phase of the first blade; is the blade rotation rate of the rotor target, is the angular velocity of rotation.

3. A method for extracting micro-motion target features with multi-domain parameter estimation and phase compensation according to claim 2, characterized in that, When calculating the peak number of echoes of rotor-like targets, for a rotor target with the number of blades being there will be a moment when peaks exist simultaneously on different blades. Therefore, the peak frequency in the time domain is not necessarily , then the number of blades is analyzed by dividing them into odd and even blades: When the number of blades is odd, let , and at this time, the moment when the th blade appears at the peak is: （3）； （4）； Verify whether there is a moment during one cycle when the th blade and the th blade simultaneously reach their peaks. If so, it meets the , that is: （5）； The left - hand side expression of Equation (5) is an even number, while the right - hand side expression is an odd number, that is does not hold. When the rotor target is an odd - blade target, there is no moment when peaks appear simultaneously on different blades. At this time, the number of peaks appearing in one cycle is , and the time - domain peak frequency is , which is twice the spectral interval; When the number of blades is even, let , and at this time, the moment when the th blade appears at the peak is: （6）； （7）； Verify whether there are different blades at the same time when the peak values appear within one cycle. If so, it meets , that is: （8）； Equation (8) has the condition for equality, that is , at this time ; Since is half of the number of blades, then the th blade and the th blade have the same corresponding moments of time-domain peaks respectively. At this time, the number of peaks appearing in one period is , and the time-domain peak frequency is , which is the same as the spectral interval; According to Equations (5) and (8), the time interval between two peaks in the time domain satisfies: （9）。 4. A method for extracting micro - motion target features with multi - domain parameter estimation and phase compensation according to claim 3, characterized in that, The specific content of S2 is that The time domain sliding window peak search algorithm is: When the sampling frequency is , and the echo length is , the echo of the rotor target is expressed as: （10）； S201. Use the window length threshold search estimation algorithm to estimate the planned window length and the peak threshold ; S202. Perform sliding window calculation to extend the echo signal and obtain an extended signal ; （11）； At this time, peak search is performed on the extended signal, specifically: Take the first moment of the signal , and the amplitude of the signal at this corresponding moment is ; extract the sliding window signal , and the formula is: （12）； Obtain the maximum value of the sliding window signal , if at the first moment , the signal echo amplitude simultaneously satisfies the conditions and , then it is considered that a peak appears at the first moment . Record the first moment in the peak set , the first peak moment ; If the condition is not met, the first moment is not included in the peak set; S203. Perform the operation of S2 on the signal moments in sequence, and finally obtain a peak set: （13）； To search for the number of moments that meet the conditions, that is, the number of peaks in the judgment, namely to complete the time-domain peak sliding window search.

5. A micro-motion target feature extraction method for multi-domain parameter estimation and phase compensation according to claim 4, characterized in that S2 also includes: The frequency domain sliding window peak search algorithm is: Perform moving target indication (MTI) operation on the rotor target echo to filter out the frequency components near zero frequency. The length of the echo is After performing MTI on the echo, perform a fast Fourier transform to obtain the frequency domain as follows: （14）； S21. Estimate the planned window length using the window length threshold search estimation algorithm and the peak threshold ; S22. Perform sliding window calculation to extend the echo signal and obtain an extended signal ; （15）； At this time, peak search is performed on the extended signal, and the steps are: Obtain the frequency of the first frequency point , and the amplitude of the corresponding moment signal at this time is , extract the sliding window signal which is （16）； Obtain the maximum value of the sliding window signal , if the frequency point , the signal echo amplitude simultaneously satisfies the conditions and , then the frequency point is considered to have a peak, and the frequency point is recorded in the peak set , the first peak frequency point , if the conditions are not satisfied, then the frequency point is not included in the peak set; S23. Perform the operation of S22 on the signal moments in sequence, and finally obtain a peak set: （17）； To search for the number of moments that meet the conditions, i.e., the number of peaks in the judgment, the frequency-domain sliding window peak search is completed.

6. The method for extracting micro-motion target features with multi-domain parameter estimation and phase compensation according to claim 5, wherein The specific content of S3 is as follows: The time-domain peak set and the frequency-domain peak set are obtained through sliding window peak search. The time-domain peak number and the frequency-domain spectral interval are obtained by using the peak set. Since there is noise and clutter in the measured data, the time difference between two adjacent time-domain peaks in the time-domain peak set is calculated. The number of time differences is 1 less than the peak number. Then, the reciprocal of the time difference is taken to obtain the time-domain peak frequency set : （18）； Set a certain acceptance estimation error range, and the final number of time-domain peaks is averaged for the acceptable results to obtain the estimated time-domain peak frequency ; For obtaining the number of frequency-domain peaks, by calculating the frequency difference between two adjacent frequency-domain peaks in the frequency-domain peak set the number of frequency differences is 1 less than the number of peaks. Then, taking the time difference as the frequency-domain spectral interval, the frequency-domain spectral interval set is obtained as follows : （19）； Set a certain acceptance estimation error range, and finally the frequency-domain spectral interval is averaged for the acceptable results to obtain the estimated spectral interval ; Estimate the peak frequency in the time domain by comparison and the estimated spectral interval to determine the parity of the number of target blades. If , the target is an odd-numbered blade; if the target is an even-numbered blade, that is, the target speed and the number of blades can be roughly estimated.

7. A micro-motion target feature extraction method for multi-domain parameter estimation and phase compensation according to claim 6, characterized in that The specific content of S4 is as follows: For a rotor target, the echo is generated by superimposing the echoes of multiple blades with different initial phases, that is, each blade echo is divided into two parts , :[[]]END]] （20）； Perform short-time Fourier transform on the signal, convert the time domain signal product into a frequency domain convolution, and the transformation result is: （21）； Perform phase compensation on the signal part, and construct a phase compensation factor : （22）； Among them, is the variable value in the compensation factor, and ； （23）； The compensated rotor target echo signal is: （24）； The compensated echo is compared with the uncompensated echo, and the compensated echo is also divided into two parts: , : （25）。 8. A method for extracting micro-motion target features with multi-domain parameter estimation and phase compensation according to claim 7, characterized in that: When the number of blades is odd, let , and the compensated echo at this time is: （26）； As known from Equation (26), only when at this time , the phase compensation of the first blade is realized. For other blades, compared with before the phase compensation and are exactly the same. At this time, the changed is manifested in the time-frequency result as: The sine amplitude changes from to ; The phase of the sine curve changes from to ; After phase compensation The time-frequency result is convolved with the time-frequency result, and the time-frequency results of each blade obtained are significantly different from those before phase compensation; The moment when each blade flickers remains unchanged, while the amplitude of the sine curve is affected. The amplitudes of the two sine curves corresponding to the th blade are as follows: （27）。 9. A method for extracting micro-motion target features with multi-domain parameter estimation and phase compensation according to claim 8, characterized in that: When the number of blades is even, let , and the compensated echo at this time is: （28）； As can be seen from Equation (28), let , at this time , and the phase compensation of the first blade is achieved at this time; when the number of target blades of the rotor is even, at each moment when there is flicker, there will be two blades flickering simultaneously. At this time, one blade has positive-frequency flicker and one blade has negative-frequency flicker. The two blades 、 satisfy ; When then （29）； When then （30）。 10. A method for extracting micro-motion target features with multi-domain parameter estimation and phase compensation according to claim 9, characterized in that: After phase compensation The time-frequency result is convolved with the time-frequency result, and the time-frequency result of each blade obtained has obvious differences from that before phase compensation. The moment when each blade flickers changes. The amplitudes of the two sine curves corresponding to the th blade are as follows: （31）； At this time, except for and the blades, the amplitudes and initial phases of the two curves in the time-frequency results of other blades are different, and the flashing frequency band is not symmetric about the zero frequency of 0; Blade , the amplitudes of the two corresponding sine curves in the time-frequency result are both , The amplitudes of the two sine curves corresponding to the blade are: （32）； In the time-frequency result, the flicker frequency band is located between two sine curves, and the amplitudes of the two curves are and , and the two sine curves reach their maximum and minimum values simultaneously; while at this time and when the time-frequency results are superimposed, the amplitudes of the two sine curves are: （33）； The flicker of the two blades is superimposed in the time-frequency result, and the two curves reach the extreme values simultaneously. When one sine curve reaches the maximum value, the other sine curve reaches the minimum value. At this time, the flicker frequency band width is A , ranging from , and the ratio of the flicker frequency band width of the positive half-axis to that of the negative half-axis is approximately .

Images