A short pulse micro-Doppler feature extraction method and device

Abstract

The present invention belongs to the field of signal processing technology, and discloses a short pulse micro-Doppler feature extraction method and device, the method comprising: according to the setting of the sonar system, dividing the target echo signal into multiple frames of short pulse micro-Doppler signals according to the pulse period; extracting the effective pulse echo range of each frame of the short pulse micro-Doppler signal; using the instantaneous frequency extraction algorithm, estimating the instantaneous frequency characteristics of the effective pulse echo range of each frame of the short pulse micro-Doppler signal. The scheme of the embodiment of the present invention converts the short pulse micro-Doppler feature extraction problem into the micro-Doppler feature extraction problem in the effective pulse range of each frame by extracting the effective pulse echo range of each frame of the signal, so that the existing long pulse micro-Doppler feature algorithm can take effect, and provides a short pulse micro-Doppler feature extraction method.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and in particular to a short pulse micro-Doppler feature extraction method and device. Background Art

[0002] In the sonar system, when the target size is much larger than the wavelength of the sound wave, the underwater target can be regarded as a collection of discrete point scatterers. In addition to the motion consistent with the translation of the target particle, these point scatterers also have periodic motion relative to the target particle, such as rotation, vibration, nutation, etc. These relative motions of the target particle are called micro-motions of the target. It is well known that the translational velocity of the target particle relative to the sonar system will modulate the detection signal, which is the Doppler effect. Similarly, the micro-motion velocity of the point scatterers on the target will also modulate the detection signal, which is the micro-Doppler effect. The signal characteristics generated by the micro-Doppler effect are called micro-Doppler characteristics, and the signal containing the micro-Doppler characteristics is called a micro-Doppler signal.

[0003] The micro-Doppler effect can be regarded as the micro-Doppler effect of the radial velocity of the target in micromotion, that is, the Doppler effect of the micromotion velocity projected on the sonar detection direction (Light of Sight, LOS). Since the angle between the motion of the point scatterer relative to the target particle and the LOS changes relatively quickly with time and often presents a periodic reciprocating nature, the modulation of the detection signal by the micro-Doppler effect makes the micro-Doppler signal often appear in the form of a periodic nonlinear frequency modulation signal, that is, the instantaneous frequency of the signal is a periodic signal or an approximately periodic signal. According to the ratio of the length of the sonar detection pulse to the period of the target micromotion, the micro-Doppler signal can be divided into a micro-Doppler signal under long pulse conditions and a micro-Doppler signal under short pulse conditions. The pulse length of the former is greater than the micromotion period of the target (generally more than twice), and the pulse length of the latter is much smaller than the micromotion period of the target.

[0004] Existing micro-Doppler feature extraction methods are mainly used for processing long-pulse micro-Doppler signals. Different from long-pulse micro-Doppler signals, there is an interval between each frame of the short-pulse micro-Doppler signal within a period of target micro-motion, and due to the micro-motion of the target, the intervals are inconsistent, which makes the time domain waveform of the echo signal truncate and the instantaneous frequency discontinuous, resulting in the failure of the signal decomposition method based on smoothness and the micro-Doppler feature based on instantaneous frequency for long-pulse micro-Doppler feature extraction methods. Therefore, a short-pulse micro-Doppler feature extraction method is needed. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the prior art, and a short pulse micro-Doppler feature extraction method is proposed. The present invention also discloses a short pulse micro-Doppler feature extraction device.

[0006] A short pulse micro-Doppler feature extraction method, comprising:

[0007] Step 1: According to the settings of the sonar system, the target echo signal is divided into multiple frames of short-pulse micro-Doppler signals according to the pulse period;

[0008] Step 2: Extract the effective pulse echo range of each frame of short pulse micro-Doppler signal;

[0009] Step 3: Use the instantaneous frequency extraction algorithm to estimate the instantaneous frequency characteristics of the effective pulse echo range of each frame of short-pulse micro-Doppler signal.

[0010] In an improved short pulse micro-Doppler feature extraction method, it also includes:

[0011] Step 4: Using the physical model of the micro-Doppler signal, the instantaneous frequency features between frames are regrouped according to the relationship between the components to obtain a set of instantaneous frequencies with uneven distribution in the time domain.

[0012] In an improved short pulse micro-Doppler feature extraction method, it also includes:

[0013] Step 5: Using the frequency estimation algorithm with non-uniform sampling time, the instantaneous frequency with non-uniform time domain distribution is processed to obtain the micro-motion frequency of the short-pulse micro-Doppler signal component.

[0014] In an improved short pulse micro-Doppler feature extraction method, step 2 includes:

[0015] Step 2-1: transform the short pulse micro-Doppler signal into the frequency domain using short-time Fourier transform;

[0016] Step 2-2: Detect the maximum value of the sub-band at each moment. When the maximum value of the sub-band is greater than the set threshold, determine that the moment is a valid pulse interval.

[0017] In an improved short pulse micro-Doppler feature extraction method, step 4 includes:

[0018] Step 4-1: Estimate the instantaneous frequency of the source attribution number {i(0) i(1) ... i(N)} at time t The frequency closest to the carrier frequency f c The estimate of is taken as the instantaneous frequency of the particle translation component and removed;

[0019] Step 4-2: Separate the outer envelopes of the instantaneous frequency estimates one by one by extracting the maximum values ​​of the results obtained in step 4-2 one by one;

[0020] Step 4-3: Regroup the instantaneous frequencies after separating the outer envelope to obtain a group of instantaneous frequencies with uneven distribution in the time domain.

[0021] The present invention also provides a short pulse micro-Doppler feature extraction device, comprising:

[0022] A signal input module is used to divide the target echo signal into multiple frames of short pulse micro-Doppler signals according to the pulse period according to the setting of the sonar system;

[0023] A single frame detection module, used to extract the effective pulse echo range of each frame of short pulse micro-Doppler signal; and

[0024] The single-frame instantaneous frequency extraction module is used to estimate the instantaneous frequency characteristics of the effective pulse echo range of each frame of short-pulse micro-Doppler signal by using the instantaneous frequency extraction algorithm.

[0025] In an improved short pulse micro-Doppler feature extraction device, it also includes:

[0026] The multi-frame regrouping module is used to utilize the physical model of the micro-Doppler signal to regroup the instantaneous frequency characteristics between frames according to the relationship between the components, and obtain a group of instantaneous frequencies with uneven distribution in the time domain.

[0027] In an improved short pulse micro-Doppler feature extraction device, it also includes:

[0028] The non-uniform frequency estimation module is used to process the instantaneous frequency with non-uniform time domain distribution using a frequency estimation algorithm with non-uniform sampling time, and obtain the micro-motion frequency of the short-pulse micro-Doppler signal component.

[0029] In an improved short pulse micro-Doppler feature extraction device, a single frame detection module includes:

[0030] A time-frequency transform unit, used for transforming the short-pulse micro-Doppler signal into the frequency domain by using a short-time Fourier transform; and

[0031] The detection unit is used to detect the maximum value of the sub-band at each moment, and when the maximum value of the sub-band is greater than a set threshold, the moment is determined to be a valid pulse interval.

[0032] In an improved short pulse micro-Doppler feature extraction device, the multi-frame regrouping module specifically includes:

[0033] The translational component separation unit is used to estimate the instantaneous frequency of the source belonging to the sequence {i(0) i(1) ... i(N)} at the estimated time t. The frequency closest to the carrier frequency f c The estimate of is taken as the instantaneous frequency of the particle translation component and removed;

[0034] An envelope separation unit, used to extract the maximum values ​​of the results obtained by the translational component separation unit one by one, and separate the outer envelopes of the instantaneous frequency estimates one by one; and

[0035] The regrouping unit is used to regroup the instantaneous frequencies after the outer envelope is separated to obtain a group of instantaneous frequencies with uneven time domain distribution. Compared with the prior art, the advantages of the present invention are:

[0036] 1. By extracting the effective pulse echo range of each frame signal, the short pulse micro-Doppler feature extraction problem is transformed into the micro-Doppler feature extraction problem in the effective pulse range of each frame, so that the existing long pulse micro-Doppler feature algorithm can take effect;

[0037] 2. Based on the physical model of micro-Doppler signals and multi-frame combination, the problem of uncertain inter-frame ordering of short-pulse micro-Doppler signals is solved;

[0038] 3. Use the frequency estimation method with non-uniform sampling time to calculate the time non-uniform sample f n The changing frequency is the micro-motion frequency of the signal component, which solves the error problem of the micro-motion frequency estimated by the frequency estimation algorithm based on uniform sampling due to the uneven time domain distribution of the instantaneous frequency of the short-pulse micro-Doppler signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a geometric relationship diagram of a rotating point target in the sonar coordinate system of the present invention;

[0040] FIG2( a ) is a time domain waveform simulation under short pulse conditions in Example 1 of the present invention;

[0041] FIG2( b ) is a simulation diagram of the time-frequency spectrum characteristics under short pulse conditions in Example 1 of the present invention;

[0042] Figure 3 is a flow chart of the short pulse Doppler feature extraction method in Example 1 of the present invention;

[0043] FIG4( a ) is a time domain waveform simulation diagram of a signal after detection in Example 1 of the present invention;

[0044] FIG4( b ) is a simulation diagram of the time-frequency spectrum characteristics of the signal after detection in Example 1 of the present invention;

[0045] FIG4( c ) is a simulation diagram of the instantaneous frequency extracted by VNMCD in Example 1 of the present invention;

[0046] FIG4( d ) is a simulation diagram of the instantaneous frequency extracted by ICCD in Example 1 of the present invention;

[0047] Figure 5 is a flow chart of the short pulse Doppler feature extraction method in Example 2 of the present invention;

[0048] FIG6( a ) is a simulation diagram of instantaneous frequency before regrouping in Example 2 of the present invention;

[0049] FIG6( b ) is a simulation diagram of instantaneous frequency after regrouping in Example 2 of the present invention;

[0050] Figure 7 It is a structural schematic diagram of a short pulse Doppler feature extraction device in Example 3 of the present invention;

[0051] Figure 8 A spatial relationship diagram between the sonar system and the target provided in Example 4 of the present invention;

[0052] Fig. 9 This is a performance comparison diagram of various algorithms in Example 4 of the present invention under different InSNR conditions;

[0053] FIG10( a ) is a time domain waveform diagram of measured data in Example 4 of the present invention;

[0054] FIG10( b ) is a time-frequency spectrum characteristic diagram of measured data in Example 4 of the present invention;

[0055] FIG10( c ) is a time-frequency spectrum characteristic diagram of a valid pulse segment of measured data in Example 4 of the present invention;

[0056] FIG10( d ) is a graph showing the instantaneous frequency characteristics of a single frame extracted in Example 4 of the present invention;

[0057] FIG10( e ) is a graph showing the instantaneous frequency characteristics of the single frame extracted after regrouping in Example 4 of the present invention;

[0058] FIG10( f ) is a time-frequency spectrum diagram of component 2 after non-uniform Fourier transformation in embodiment 4 of the present invention. DETAILED DESCRIPTION

[0059] In order to better understand the scheme of the present invention, the truncation effect of the time domain waveform of the short pulse micro-Doppler signal is first explained below, and then the short pulse micro-Doppler signal model is described, followed by a description of the inventive concept of the present invention, and finally a specific embodiment is given.

[0060] 1) The time domain waveform of the short pulse micro-Doppler signal has a truncation effect: According to the simulation parameters shown in Table (1), the numerical simulation of the simplest short pulse multi-component micro-motion target system consisting of a translation point and a rotation point, that is, the reflected echo of the target rotating around the axis, is given. The pulse interval is taken as the sampling interval in the slow time domain, the signal within a pulse interval is called a frame signal, and the time index of the sampling data in each frame is taken as the time index in the fast time domain. The time domain waveform of the signal is stacked on the slow time-fast time plane, and the time domain waveforms of the signal reflected echo of the target rotating around the axis under the short pulse condition are shown in Figure 2(a). The length of a frame signal is taken as the short time window length and the frame shift length, and the frequency domain characteristics of the signal reflected echo of the target rotating around the axis under the short pulse condition are shown in Figure 2(b).

[0061] Table (1) Simulation parameter settings under short pulse conditions

[0062]

[0063] 2) Short pulse micro-Doppler signal model

[0064] Since the target reflection echo in actual detection can often be equivalent to the geometry of a discrete point scatterer, the micro-Doppler signal model of any target can be expressed as the sum of multiple single-point target signal models. The short-pulse micro-Doppler signal can be regarded as sampling the long-pulse Doppler signal using a time window related to the relative distance between the target and the sonar. Here, the ideal single-point target rotation signal model is first introduced in 21) below, and the simplified conditions of the underwater acoustic channel are introduced in 22) below and the time window of the short-pulse signal is derived to give the ideal short-pulse micro-Doppler signal model.

[0065] 21) Micro-Doppler signal model of point target rotation

[0066] In order to establish an ideal signal model for single-component micro-Doppler signals, we first assume a single target that performs independent micro-movements in three-dimensional space. Assume that the target is a rigid body, the target scale satisfies the conditions of a point scattering target, and the target performs uniform rotation around a fixed point in space. The target coordinate system o′x′y′z′ is established with an arbitrary point in the space as the origin o′, and the sonar detection center is used as the origin o to establish the sonar coordinate system oxyz. The geometric relationship is as follows: Figure 1 shown.

[0067] According to the geometric relationship shown in Figure (1), for any time t, the position vector d(t) (red) of a point P rotating around a fixed axis in space in the sonar coordinate system oxyz can be expressed as the position vector r of the target coordinate system relative to the center of the sonar system s (blue), the displacement vector vt at time t, and the position vector r1(t) (blue) in the target coordinate system o′x′y′z′ at time t, that is:

[0068]

[0069] Among them, r0 (green) is the position vector of the rotation center of the point target P in the target coordinate system o′x′y′z′, and r (red) is the position vector of point P relative to the rotation center. Ω (t) is the rotation matrix of the position vector r to r′(t).

[0070] At this time, the central frequency of the sonar system is f c The echo of the target point P can be expressed as a function of the time t and its position d(t) in the sonar coordinate system:

[0071]

[0072] Where σ is the reflection coefficient of the target, c is the underwater sound speed, and d(t) = |d(t)| is the distance of the rotating point target relative to the center of the sonar system.

[0073] Compared with the propagation signals of electromagnetic waves and ultrasonic waves in space, the propagation channel of underwater sound waves is more complicated. In order to establish an ideal signal model, it is necessary to introduce two assumptions that reasonably simplify the underwater acoustic channel: free sound field conditions and far field conditions. Assume that the target and the sonar system are completely submerged in the water and the distance between them meets the following requirements: 1. It is much larger than the distance between the target and the sonar system and the horizontal plane and the bottom of the water respectively, so the multi-channel components caused by the water surface and the bottom of the water can be ignored, that is, the free sound field condition; 2. It is much larger than the distance of the target point relative to the origin of the target coordinate system, that is, the geometric structure of the target can be ignored and it is equivalent to a point target located in the sonar-target direction, that is, the far field condition is satisfied. s 》r1(t). Therefore, the single-point rotation micro-Doppler signal model shown in equation (2) can be approximately expressed as shown in equation (3).

[0074]

[0075] In which, it is assumed that the target is completely reflective, that is, the reflection coefficient of the target is σ = 1, r s =|r s | is the initial distance between the sonar and the target, v is the relative radial velocity between the sonar and the target, ω=|ω| is the angular velocity of the point target P around the fixed axis, r e =rcosβ is the equivalent rotation axis length in the LOS radial plane, r = |r| is the distance of the point target P relative to the rotation center, that is, the rotation axis length, β is the angle between the rotation axis and LOS, It is the angle between the rotation angular velocity ω and the rotation axis r at time t=0.

[0076] 22) Short pulse micro-Doppler signal model

[0077] According to the single-point signal model derived in 21), the signal model of multi-component micro-Doppler signal under short pulse condition is given.

[0078] According to the point scattering model, the micro-Doppler signal model of any target can be expressed as the sum of multiple single-point target signal models. Therefore, a multi-component target can be expressed as the reflected echo s0(t) of the target particle and the reflected echo s of the micro-motion point on the target. n (t), the sum of n = 1, 2...N, which is the form shown in formula (4).

[0079]

[0080] The short pulse condition means that within a pulse repetition period T_interval, the pulse width T_pulse is much smaller than the micro-motion period T mD =2π / ω. Compared with the micro-Doppler signal of a single-point target under long pulse conditions, the target instantaneous frequency characteristics under short pulse conditions are in a time-inhomogeneous state. The interval of the target instantaneous frequency characteristics in the effective pulse echo time domain of each pulse cycle is the sampling interval T s =1 / fs, and there is no instantaneous frequency feature of the target in other time domains. Therefore, the short pulse micro-Doppler signal model can be expressed as Equation (4) in the form of non-uniform sampling in the time window related to the target distance, as shown in Equation (5).

[0081]

[0082] Where P is the number of pulses within the observed signal range, w n,p (t) is the square window of the nth component in the pth frame, which is only The pulse width T_pulse is much smaller than the micro-motion period T under the short pulse condition. mD =2π / ω, so the nth component is at the beginning of the effective pulse of the pth frame and the nth component at the end of the effective pulse of the pth frame can be approximated as and

[0083] 3) The inventive concept of the present invention: Due to the discontinuous effect of short pulse signals, the existing VNCMD (Variational Non-linear Chirp Mode Decomposition) type algorithms based on smoothness constraints and ICCD (Intrinsic Chirp Component Decomposition) type algorithms based on time-frequency spectrum ridge paths fail between frames. In order to solve this problem and obtain the instantaneous frequency of short pulse micro-Doppler signals, the present invention first performs single-frame signal detection on each frame signal to obtain the effective pulse echo range, and then performs single-frame instantaneous frequency extraction on each frame signal, that is, only extracts the instantaneous frequency features within the effective pulse echo range, and realizes the conversion of the micro-Doppler feature extraction problem of short pulse echo into the micro-Doppler feature extraction problem within the effective pulse range of each frame, so that the existing algorithm can take effect and realize the extraction of short pulse micro-Doppler features.

[0084] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0085] Example 1

[0086] like Figure 3 As shown, it is a flow chart of the short pulse micro-Doppler feature extraction method of an embodiment of the present application, comprising the following steps:

[0087] Step 1: According to the pulse period T_interval set by the sonar system, the target echo signal x(t) is divided into multiple frames of short pulse micro-Doppler signals.

[0088] Among them, the short pulse Doppler signal of the pth frame is the micro-Doppler signal echo s of each component n (t),n=1,2,...N according to the window function w n,p (t) The sum of the samples, that is,

[0089] Step 2: Extract the effective pulse echo range of each frame of short pulse micro-Doppler signal.

[0090] In this step 2, the sub-band maximum value detection method can be used to extract the effective pulse echo range, or CAFR (Constant False-Alarm Rate, constant false alarm detection), energy entropy ratio detection, power detection, information entropy detection, etc. can be used to detect the effective pulse echo range.

[0091] When the sub-band maximum value detection method is used for extraction, the following steps 2-1 and 2-2 may be included:

[0092] Step 2-1: transform the short pulse micro-Doppler signal into the frequency domain using a short-time Fourier transform with a frame shift of 1;

[0093] The time-frequency domain amplitude X(k,f) is:

[0094]

[0095] Wherein, k=1,2,...,N is the time index in the time-frequency domain, and g(w) is the time window function such as Hanning window, Hamm window, etc.

[0096] Here, we can use short-time Fourier transform with a frame shift greater than 1 to transform the short-pulse micro-Doppler signal into the frequency domain. If short-time Fourier transform with a frame shift of 1 is not applicable, interpolation is required. In addition to short-time Fourier transform, Wigner-Wiley transform, wavelet transform and other time-frequency transforms, we can also transform the short-pulse micro-Doppler signal into the frequency domain.

[0097] Step 2-2: For the maximum value of each subband f at each moment A detection is performed, and when the sub-band maximum value is greater than the set threshold, the moment is determined to be a valid pulse interval.

[0098] Specifically, for the p-th frame signal x p (t), the amplitude of the fth frequency band in the kth time frame of the time spectrum with a frame shift of 1 is expressed as |X p (k,f)|, extract the effective pulse echo range t∈K of the p-th frame signal p As shown in formula (6).

[0099] K p ={k||X p (k,f)|≥αmax(|X p (k,f)|)} (6)

[0100] Wherein, α is an amplitude factor, and in this embodiment, the amplitude factor α=0.7 is selected.

[0101] Step 3: Use the instantaneous frequency extraction algorithm to estimate the instantaneous frequency characteristics of the effective pulse echo range of each frame of short-pulse micro-Doppler signal.

[0102] In this step 3, the existing instantaneous frequency extraction algorithm based on modal decomposition can be used to estimate the instantaneous frequency characteristics of the effective pulse echo range of each frame of short pulse micro-Doppler signal, or the instantaneous frequency extraction algorithm based on VNCMD, ICCD, and adaptive time-frequency basis component separation can be used to estimate the instantaneous frequency characteristics of the effective pulse echo range of each frame of short pulse micro-Doppler signal.

[0103] The present invention adopts an instantaneous frequency extraction algorithm based on modal decomposition to estimate the instantaneous frequency characteristics of the effective pulse echo range of each frame of short-pulse micro-Doppler signal.

[0104] Specifically, for the multi-component micro-Doppler signal x(t), the estimation of the instantaneous frequency is In the case of , the signal modal decomposition method based on smoothness constraint and based on eigenmodel constraint pair can be expressed as shown in Equation (7) and Equation (8) respectively.

[0105]

[0106] In formula (7), a n ={a n (t1) a n (t2) ... a n (t T )} is the instantaneous amplitude of the signal, T and △t are the number of samples and sampling interval of the discrete signal respectively, H n Based on the instantaneous frequency estimate The data matrix of Ω is the second-order differential operator matrix, μ is the Lagrangian operator, and f n (t) is the actual value of the instantaneous frequency.

[0107]

[0108] In formula (8), is the model parameter of the instantaneous amplitude in the eigenvalue model, ω0 is the fundamental frequency of the Fourier basis function, K1 is the model order of the instantaneous amplitude, Φ n Based on the instantaneous frequency estimate The data matrix, Z n is based on the eigenmodel basis function (Fourier basis function) and Φ n is the data matrix of , and μ is the Lagrangian operator.

[0109] The instantaneous frequency estimation It can also be estimated by the inverse tangent model based on the instantaneous amplitude bandwidth minimum principle and the intrinsic mode model based on the time-frequency spectrum ridge path, as shown in equations (9) and (10).

[0110]

[0111] Where X(t,f) is the time-frequency spectrum of the signal. Combining equation (7) with equation (9) gives the VNCMD algorithm. Combining equation (8) with equation (10) and setting It is expressed as the sum of finite-order Fourier basis functions, as shown in formula (11), which is the ICCD algorithm.

[0112]

[0113] in, is the model parameter of the instantaneous frequency in the eigenmodel, and K2 is the model order of the instantaneous frequency.

[0114] Taking FIG4(a) and FIG4(b) as examples, the instantaneous frequency extraction results of the detected signal using VNMCD and ICCD are shown in FIG4(c) and FIG4(d).

[0115] Example 2

[0116] There is an interval between each frame of echoes of the short-pulse micro-Doppler signal, and due to the micro-motion of the target, the intervals are different, which causes a truncation effect on the time domain waveform of the echo signal and discontinuous instantaneous frequency, resulting in failure of the signal decomposition method based on smoothness and the micro-Doppler feature extraction method based on instantaneous frequency. Embodiment 1 of the present invention provides a solution. However, there is an interval between each frame of echoes of the short-pulse micro-Doppler signal, and due to the micro-motion of the target, it also leads to a larger time interval and uncertain ordering problems in the signal decomposition algorithm. The existing algorithm cannot distinguish the source attribution of each component, which causes the extracted micro-Doppler features to have uncertain ordering problems between frames. In order to solve the problem of uncertain ordering between frames of the extracted micro-Doppler features, Embodiment 2 of the present invention provides a short-pulse micro-Doppler feature extraction method, such as Figure 5 As shown, the following steps are included:

[0117] Step 1: According to the settings of the sonar system, the target echo signal is divided into multiple frames of short-pulse micro-Doppler signals according to the pulse period.

[0118] Step 2: Extract the effective pulse echo range of each frame of short pulse micro-Doppler signal.

[0119] Step 3: Use the instantaneous frequency extraction algorithm to estimate the instantaneous frequency characteristics of the effective pulse echo range of each frame of short-pulse micro-Doppler signal.

[0120] Step 4: Using the physical model of the micro-Doppler signal, the instantaneous frequency features between frames are regrouped according to the relationship between the components to obtain a set of instantaneous frequencies with uneven distribution in the time domain.

[0121] Here, the instantaneous frequency features between frames are regrouped, so that the short pulse micro-Doppler feature extraction method of the present invention is a short pulse micro-Doppler feature extraction method based on multi-frame combination.

[0122] Using the physical model of micro-Doppler signals, short-pulse micro-Doppler signals can be expressed as multi-component micro-Doppler signals. The instantaneous frequencies are expressed by formulas (12) and (13).

[0123]

[0124] in, is the micro-Doppler frequency shift of n components, v is the relative radial velocity between the sonar and the target, c is the underwater sound speed, t is any time, f c is the carrier frequency.

[0125] The relative carrier frequency f between different components can be obtained by using the instantaneous frequency of the signal components shown in equations (12) and (13): c The relationship between the two is relatively fixed, and is closest to the carrier frequency f c The component is the component of the particle translation, so this step 4 can be specifically implemented by following the steps 4-1 to 4-3.

[0126] Step 4-1: Estimate the instantaneous frequency of the source attribution number {i(0) i(1) ... i(N)} at time t The frequency closest to the carrier frequency f c An estimate of is taken as the instantaneous frequency of the particle translational component and removed.

[0127] Step 4-2: Extract the maximum values ​​of the results obtained in step 4-1 one by one, that is, separate the outer envelopes of each instantaneous frequency estimate one by one.

[0128] The mathematical expressions of the above steps 4-1 and 4-2 can be shown as formula (14) and formula (15).

[0129]

[0130] Where i(m n ),n=1,...,N is the source attribution number of the corresponding result.

[0131]

[0132] Where i(m n ),n=1,...,N is the source attribution number of the corresponding result.

[0133] Step 4-3: Regroup the instantaneous frequencies after separating the outer envelope using the existing regrouping method to obtain a set of instantaneous frequencies f with uneven distribution in the time domain. n ={f n,1 f n,2 ... f n,P}, where f n,p ={f n,p (1) f n,p (2)... f n,p (T p )} is the instantaneous frequency vector of the p-th frame signal of the n-th component.

[0134] As shown in Figure 6(a), the instantaneous frequency diagram after single-frame signal detection and extraction of each frame of short-pulse micro-Doppler signal has obvious inter-frame order uncertainty. In Figure 6(a), component 1 is the micro-motion frequency component of the short-pulse micro-Doppler signal, and component 2 is the translational frequency component of the short-pulse micro-Doppler signal. According to the scheme of this implementation 2, the instantaneous frequency after regrouping is shown in Figure 6(b), which basically eliminates the inter-frame order uncertainty.

[0135] The above method proposed in this embodiment is a multi-frame joint regrouping method based on the physical model of micro-Doppler signals. It can be expressed as separating the components of particle translation through equation (14), and then separating the outer envelopes of each instantaneous frequency estimate one by one through equation (15). The separated outer envelopes can be regrouped through the existing regrouping method to solve the problem of uncertain inter-frame ordering.

[0136] In the extraction of micro-Doppler features, in addition to extracting the instantaneous frequency of each component, it is also necessary to estimate the micro-motion frequency of each component. Since the time domain distribution of the instantaneous frequency of the short pulse micro-Doppler signal is uneven, the micro-motion frequency estimated by the frequency estimation algorithm based on uniform sampling has errors. In order to compensate for this part of the error, preferably, the method of the present invention also includes:

[0137] Step 5: Using the frequency estimation algorithm with non-uniform sampling time, the instantaneous frequency with non-uniform time domain distribution is processed to obtain the micro-motion frequency of the short-pulse micro-Doppler signal component.

[0138] Using the frequency estimation method of non-uniform sampling time to calculate the time non-uniform samples f n The frequency of change is the micro-vibration frequency of the signal component.

[0139] For non-uniform instantaneous frequency samples f n , let the corresponding time be t={t1 t2 ... t P}, where the instantaneous frequency of the p-th frame signal corresponds to the time T p is the number of samples extracted from the p-th frame signal. Then the instantaneous frequency sample at any time can be expressed as f n,p (i) = f n (t p,i ), the non-uniform Fourier transform can be expressed as shown in formula (16). The frequency point corresponding to the peak value of the signal frequency domain after the non-uniform Fourier transform is the micro-motion frequency ω of the target component. n , as shown in formula (17).

[0140]

[0141] Among them, ave(.) is the mean operator and T is the duration of the observed signal.

[0142] In the scheme of the embodiment of the present invention, according to the changing characteristics of the instantaneous frequency of the micro-Doppler signal, a feature extraction method suitable for short-pulse micro-Doppler signals is proposed by utilizing a slowly varying Fourier basis and a fractional-order trigonometric function basis based on frequency estimation. This method improves the estimation accuracy of the instantaneous frequency characteristics of the multi-component micro-Doppler signal and completes the estimation of the micro-motion period of multiple targets.

[0143] Example 3

[0144] Embodiment 3 of the present invention provides a short pulse micro-Doppler feature extraction device, such as Figure 7 As shown, it includes: a signal input module 71, a single frame detection module 72 and a single frame instantaneous frequency extraction module 73; wherein:

[0145] The signal input module 71 is used to divide the target echo signal into multiple frames of short pulse micro-Doppler signals according to the pulse period according to the setting of the sonar system;

[0146] A single frame detection module 72, used to extract the effective pulse echo range of each frame of short pulse micro-Doppler signal;

[0147] The single-frame instantaneous frequency extraction module 73 is used to estimate the instantaneous frequency characteristics of the effective pulse echo range of each frame of short-pulse micro-Doppler signal by using an instantaneous frequency extraction algorithm.

[0148] Specifically, the single frame detection module 72 includes: a time-frequency conversion unit 721 and a detection unit 722;

[0149] A time-frequency transform unit 721 is used to transform the short pulse micro-Doppler signal into the frequency domain by using a short-time Fourier transform;

[0150] The detection unit 722 is used to detect the maximum value of the sub-band at each moment, and when the maximum value of the sub-band is greater than a set threshold, determine that the moment is a valid pulse interval.

[0151] Preferably, the short pulse micro-Doppler feature extraction device further includes:

[0152] The multi-frame regrouping module 74 is used to regroup the instantaneous frequency characteristics between frames according to the relationship between components by using the physical model of the micro-Doppler signal, so as to obtain a group of instantaneous frequencies with uneven distribution in the time domain.

[0153] Preferably, the short pulse micro-Doppler feature extraction device further includes:

[0154] The non-uniform frequency estimation module 75 is used to process the instantaneous frequency with non-uniform time domain distribution using a frequency estimation algorithm with non-uniform sampling time, and obtain the micro-motion frequency of the short pulse micro-Doppler signal component.

[0155] Preferably, the multi-frame regrouping module 74 specifically includes:

[0156] The translational component separation unit 741 is used to estimate the instantaneous frequency of the source belonging sequence {i(0) i(1) ... i(N)} at the estimated time t. The frequency closest to the carrier frequency f c The estimate of is taken as the instantaneous frequency of the particle translation component and removed;

[0157] An envelope separation unit 742 is used to extract the maximum values ​​of the results obtained by the translation component separation unit one by one, and separate the outer envelopes of each instantaneous frequency estimate one by one;

[0158] The regrouping unit 743 is used to regroup the instantaneous frequencies after the outer envelope is separated to obtain a group of instantaneous frequencies with uneven time domain distribution.

[0159] Example 4

[0160] The following is an explanation of the simulation experiment and actual test of the short pulse micro-Doppler feature extraction method proposed by the present invention through this embodiment.

[0161] Simulation experiment: Changes and comparison of algorithm performance in signal extraction as input signal-to-noise ratio changes

[0162] Generate a pure input signal according to the short pulse micro-Doppler signal simulation parameters shown in Table (1), and assume that the signal-to-noise ratio (InSNR) of the additive Gaussian noise of the input segment ranges from 0dB to 10dB with an interval of 1dB. The spatial relationship between the sonar system and the target in this experiment is as follows: Figure 8 shown.

[0163] The existing well-recognized VNCMD and ICCD are used as benchmark algorithms to compare the signal extraction performance of the two algorithms directly processing short pulse micro-Doppler signals and using the short pulse micro-Doppler feature extraction (Joint Micro-Doppler Feature Extraction, JMDFE) algorithm based on multi-frame joint proposed by the present invention. In order to obtain a smooth performance curve, a Monte Carlo random experiment is used, and the number of experiments is 100. The performance comparison of each algorithm is as follows Fig. 9 As shown in the figure. The black baseline indicates that the input and output signal-to-noise ratios are the same, that is, the dividing line of zero optimization. Obviously, the existing micro-Doppler feature extraction algorithms, whether VNCMD or ICCD, are below the baseline, that is, they cannot complete the signal extraction of the micro-Doppler features of short pulse signals. However, VNCMD and ICCD combined with the JMDFE process can effectively complete the signal extraction of the micro-Doppler features of short pulse signals.

[0164] Field experiment: Estimation of micromotion frequency from measured data

[0165] In order to test the properties of the micro-Doppler effect in the underwater acoustic environment and verify the effectiveness of the proposed short-pulse micro-Doppler feature extraction algorithm, this paper applies the proposed algorithm to actual observation data. In this measured data, in order to reduce the impact of the underwater acoustic channel, the experiment uses a 1.5m*1.0m*0.6m echo-free pool as the experimental scene, uses a small ball connected to a motor as a target to simulate the physical object corresponding to the point target rotation model, and uses a single base station active sonar as the detection system. The spatial relationship between the sonar and the target is as follows: Figure 8 As shown, the specific parameter settings of the sonar-target system are consistent with 0 except T_pulse = 0.5ms:

[0166] The time domain waveform and time-frequency domain characteristics of the measured signal after matched filtering and dimensionality reduction are shown in Figures 10(a) and (b), respectively. The signal at the fast time of 0ms to 0.5ms in Figure 10(a) is the direct signal from the sonar transmitter to the receiver. Since its appearance time is fixed and has nothing to do with the target signal, it can be directly cut off. The time-frequency spectrum shown in Figure 10(b) is used as the input of the single-frame detection. The time-frequency domain characteristics of the effective pulse segment of the signal obtained after detection are shown in Figure 10(c). The instantaneous frequency of each frame extracted by ICCD is shown in Figure 10(d). The instantaneous frequency after regrouping using the micro-Doppler model is shown in Figure 10(e). The estimation of the micro-motion frequency of component 2 is shown in Figure 10(f). After JMDFE processing, the micro-motion frequency of the target component is 13.1624Hz, and the relative error between it and the actual value of 13.2 is only 0.28%.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention is described in detail with reference to the embodiments, it should be understood by those skilled in the art that any modification or equivalent replacement of the technical solutions of the present invention does not depart from the spirit and scope of the technical solutions of the present invention and should be included in the scope of the claims of the present invention.

Claims

1. A short pulse micro-Doppler feature extraction method, comprising: Step 1: According to the settings of the sonar system, the target echo signal is divided into multiple frames of short-pulse micro-Doppler signals according to the pulse period; Step 2: Extract the effective pulse echo range of each frame of short pulse micro-Doppler signal; Step 3: Using the instantaneous frequency extraction algorithm, estimate the instantaneous frequency characteristics of the effective pulse echo range of each frame of short pulse micro-Doppler signal; Also includes: Step 4: Using the physical model of the micro-Doppler signal, regroup the instantaneous frequency features between frames according to the relationship between the components to obtain a set of instantaneous frequencies with uneven distribution in the time domain; Step 4 includes: Step 4-1: Estimate the instantaneous frequency of the source attribution number {i(0)i(1)...i(N)} at time t The frequency closest to the carrier frequency f c The estimate of is taken as the instantaneous frequency of the particle translation component and removed; Step 4-2: Separate the outer envelopes of the instantaneous frequency estimates one by one by extracting the maximum values ​​of the results obtained in step 4-2 one by one; Step 4-3: Regroup the instantaneous frequencies after separating the outer envelope to obtain a group of instantaneous frequencies with uneven distribution in the time domain.

2. The short pulse micro-Doppler feature extraction method according to claim 1, characterized in that: Also includes: Step 5: Using the frequency estimation algorithm with non-uniform sampling time, the instantaneous frequency with non-uniform time domain distribution is processed to obtain the micro-motion frequency of the short-pulse micro-Doppler signal component.

3. The short pulse micro-Doppler feature extraction method according to claim 1, characterized in that: Step 2 includes: Step 2-1: transform the short pulse micro-Doppler signal into the frequency domain using short-time Fourier transform; Step 2-2: Detect the maximum value of the sub-band at each moment. When the maximum value of the sub-band is greater than the set threshold, determine that the moment is a valid pulse interval.

4. A short pulse micro-Doppler feature extraction device, characterized in that: include: A signal input module is used to divide the target echo signal into multiple frames of short pulse micro-Doppler signals according to the pulse period according to the setting of the sonar system; Single frame detection module, used to extract the effective pulse echo range of each frame of short pulse micro-Doppler signal; and A single-frame instantaneous frequency extraction module is used to estimate the instantaneous frequency characteristics of the effective pulse echo range of each frame of short-pulse micro-Doppler signal using an instantaneous frequency extraction algorithm; Also includes: A multi-frame regrouping module is used to regroup the instantaneous frequency characteristics between frames according to the relationship between components by using the physical model of micro-Doppler signals, and obtain a set of instantaneous frequencies with uneven distribution in the time domain; The multi-frame regrouping module specifically includes: The translational component separation unit is used to estimate the instantaneous frequency of the source belonging to the sequence number {i(0)i(1)...i(N)} at the estimated time t The frequency closest to the carrier frequency f c The estimate of is taken as the instantaneous frequency of the particle translation component and removed; An envelope separation unit, used to extract the maximum values ​​of the results obtained by the translational component separation unit one by one, and separate the outer envelopes of the instantaneous frequency estimates one by one; and The regrouping unit is used to regroup the instantaneous frequencies after the outer envelope is separated to obtain a group of instantaneous frequencies with uneven distribution in the time domain.

5. The short pulse micro-Doppler feature extraction device according to claim 4, characterized in that: Also includes: The non-uniform frequency estimation module is used to process the instantaneous frequency with non-uniform time domain distribution using a frequency estimation algorithm with non-uniform sampling time, and obtain the micro-motion frequency of the short-pulse micro-Doppler signal component.

6. The short pulse micro-Doppler feature extraction device according to claim 4, characterized in that: Single frame detection module, including: A time-frequency transform unit, used for transforming the short-pulse micro-Doppler signal into the frequency domain by using a short-time Fourier transform; and The detection unit is used to detect the maximum value of the sub-band at each moment, and when the maximum value of the sub-band is greater than a set threshold, the moment is determined to be a valid pulse interval.

Images