CN110163040B - Radar radiation source signal identification technology in non-Gaussian clutter - Google Patents

Abstract

The invention belongs to the technical field of radar signal processing in a non-Gaussian clutter environment, and discloses a radar signal identification method based on generalized variational modal decomposition combined with a support vector machine, which is used for carrying out generalized variational modal decomposition on a received radar radiation source signal to obtain an intrinsic modal component IMF; calculating a smooth pseudo Wigner-Ville time frequency distribution matrix of each eigenmode component, and extracting a R é nyi entropy characteristic construction characteristic vector T of each time frequency distribution; finally, carrying out classification and identification by using a support vector machine; when the generalized signal-to-noise ratio is greater than 5dB, the recognition rate of each type of signal reaches more than 65%, and particularly when the generalized signal-to-noise ratio is greater than 10dB, the recognition rate of each type of signal reaches more than 90%, so that the recognition effect of the method is good.

Description

Radar Radiator Signal Recognition Technology in Non-Gaussian Clutter

technical field

The invention belongs to the technical field of radar signal processing, and in particular relates to a radar radiation source signal classification and identification method based on generalized variational mode decomposition in a non-Gaussian clutter environment.

Background technique

Radar emitter signal recognition is a key technology in radar electronic reconnaissance, which directly affects the performance of electronic reconnaissance equipment and is related to subsequent combat decisions. According to the basis, it has a very important position and function in the process of radar electronic countermeasures, so it is of practical significance to study the identification of radar emitter signal. At present, many achievements have been made in the research on radar emitter signal recognition in Gaussian clutter environment. Ataollah Abrahamzadeh et al. used the 8th-order moment and 8th-order cumulant for identification, which can achieve a good recognition rate under the condition of low SNR, but with the increase of cumulant and moment, the calculated value is relatively large (Abrahamzadeh A, Seyedin S A, Dehghan M. Digital-signal-type identification using an efficient identifier [J]. Eurasip Journal on Advances in Signal Processing, 2007, 2007(1): 037690.); Pu Yunwei et al proposed secondary feature extraction based on instantaneous frequency However, due to the limitations of the instantaneous frequency estimation method, this method also requires a relatively high signal-to-noise ratio (Pu Yunwei, Jin Weidong, Hu Laizhao. Radiator signal classification based on secondary feature extraction of instantaneous frequency[ J].Journal of Southwest Jiaotong University, 2007,42(3):373-379.); Prakasam uses the statistical histogram of wavelet coefficients as a feature to identify the signal, and can achieve a better signal-to-noise ratio under the condition of 5dB recognition rate, but this method is mainly aimed at communication signals (Prakasam P, Madheswaran M. Digital Modulation identification model using wavelet transform and statistical parameters [J]. Journal of Computer Systems Networks & Communications, 2008, 2008 (3): 6.); Yu Zhibin proposed a wavelet-based Packet decomposition, using energy entropy and probability entropy to form feature vectors, can obtain a relatively satisfactory correct recognition rate in a large range of signal-to-noise ratios, but this method can only achieve better recognition results in a noise environment with a moderate intensity or higher. In the case of low noise ratio (SNR<5dB), the recognition effect of many signals is not ideal (Yu Zhibin, Chen Chunxia, Jin Weidong. Radiation source signal recognition based on fusion entropy features[J]. Modern Radar, 2010, 32(1 ):34-38.); For the classification of radar signals with low probability of interception, Persson C proposes to use Wigner-Ville transform to obtain the time-frequency diagram of the signal first, and then perform image processing on it and extract features, but the experimental results show that the recognition effect of this method is It is not very ideal, mainly because the cross term of the Wigner-Ville distribution seriously affects the feature extraction of the signal. In addition, the image processing method used in this document cannot effectively reduce the influence of time-frequency image noise (Persson C.Classification and Analysis of L ow Probability of Intercept RadarSignals Using Image Processing[J].Thesis Collection,2003.); Zou Xingwen proposed to convert the time-frequency image of the radar signal into a grayscale image, and directly use the normalized pixels as the recognition features, and achieved relatively good results. Good recognition effect, but because the pixel points are directly used as features in this paper, the feature dimension is large, which is easy to cause "dimension disaster", and only 5 kinds of radar radiation source signals are classified and recognized in this paper, so the validity of more types of signals Further verification is needed (Zou Xingwen, Zhang Gexiang, Li Ming, et al. A new method for radar emitter signal classification [J]. Data Acquisition and Processing, 2009, 24(4): 487-492.). However, some spike-like clutter inevitably exists in the actual radar detection environment. The distribution characteristics of this kind of clutter are different from the Gaussian distribution, and it is usually described by α-stable distribution. Because non-Gaussian clutter does not have limited second-order and above-order moments, the existing radar emitter signal identification method in Gaussian clutter environment is no longer applicable.

To sum up, the existing technology has the following problems: it is only applicable to the environment of Gaussian clutter, and for the interference of non-Gaussian clutter, the existing technology cannot achieve good results, and it is only suitable for the identification of a few source signals , the recognition effect for multiple source signals is poor.

Contents of the invention

Aiming at the problems existing in the prior art, the present invention provides a radar radiation source signal identification method under non-Gaussian clutter.

The present invention is realized in this way, radar emitter signal identification technology in non-Gaussian clutter, the radar emitter signal identification method in non-Gaussian clutter comprises:

Step 1, performing generalized variational mode decomposition on the received signal to obtain K eigenmode components IMF;

Step 2, calculate the smooth pseudo-Wigner-Ville time-frequency distribution matrix of each eigenmode component, and extract the Rényi entropy feature of each time-frequency distribution to construct the feature vector T;

Step three, use the support vector machine for classification recognition.

Further, the process of performing generalized variational mode decomposition on the received signal to obtain K eigenmode components IMF is as follows:

1) Perform nonlinear transformation on the received signal r(t), and obtain f(t), namely:

令其初始值均为0，且将设置分解模态数为K(K为整数，此处设置为K＝10)。2) Initialization

Let their initial values be 0, and set the number of decomposition modes to K (K is an integer, here K=10).

3) n=n+1, execute the loop;

和

更新uk和ω_k；4) According to

and

update _uk and ωk;

5) Update λ, namely

结束循环，6) Given the discriminant accuracy ε, until the iteration stop condition is reached

end the loop,

及中心频率ω_k，最后由傅里叶反变换得到K个窄带IMF分量。get each

and center frequency ω _k , and finally K narrowband IMF components are obtained by inverse Fourier transform.

Further, the smooth pseudo-Wigner-Ville time-frequency distribution matrix of each eigenmode component is calculated, and the Rényi entropy feature construction feature vector T of each time-frequency distribution is extracted, and the process is as follows

1) Calculate the smooth pseudo-Wigner-Ville time-frequency transformation of the VMD modal component u _k to obtain the time-frequency distribution matrix P(t,f) of the signal;

In the formula: h(τ) and g(τ) are two real even window functions, and h(0)=g(0)=1.

2) Extract the Rényi entropy of the signal time-frequency distribution P(t,f):

α is the order number of Rényi entropy, and α=2 is set in the present invention;

3) Construct the feature vector T:

Further, the support vector machine classifier is used to classify and identify radar emitter signals in non-Gaussian clutter to realize radar emitter signal type identification.

The advantages and positive effects of the invention are as follows: using generalized variational mode decomposition to extract time-frequency distribution entropy features to identify radar radiation source signal types; the invention has better identification effect.

Description of drawings

Fig. 1 is a flow chart of a method for identifying a radar emitter signal under non-Gaussian clutter provided by an embodiment of the present invention.

Fig. 2 is a schematic diagram of radar emitter signal recognition performance under non-Gaussian clutter provided by an embodiment of the present invention.

detailed description

In order to make the object, technical solution and advantages of the present invention more clear, the present invention will be further described in detail below in conjunction with the examples. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

The application principle of the present invention will be described in detail below in conjunction with the accompanying drawings.

As shown in Figure 1, a method for identifying a radar emitter signal in non-Gaussian clutter provided by an embodiment of the present invention includes the following steps:

S101: Perform generalized variational mode decomposition on the received signal to obtain K eigenmode components IMF;

S102: Calculate the smooth pseudo-Wigner-Ville time-frequency distribution matrix of each eigenmode component, and extract the Rényi entropy features of each time-frequency distribution to construct a feature vector T;

S103: Use the support vector machine to perform classification recognition.

The application principle of the present invention will be further described below in conjunction with the accompanying drawings.

A method for identifying a radar emitter signal in non-Gaussian clutter provided by an embodiment of the present invention includes the following steps:

S1 performs generalized variational mode decomposition on the received signal to obtain K eigenmode components IMF;

Received signal model under non-Gaussian clutter, its expression is:

r(t)=s(t)+w(t);

Among them, s(t) is the transmitted radar signal, conventional pulse signal (CW), linear frequency modulation (LFM) signal, phase encoding (PSK) signal, frequency encoding (FSK) signal, even quadratic frequency modulation (EQFM) signal, FM Continuous wave (FMCW) signal, COSTAS frequency modulation signal. w(t) is non-Gaussian correlated clutter.

If s(t) is a conventional pulse signal (CW), its expression is:

Among them: A is the signal amplitude, f ₀ is the carrier frequency, T is the pulse width.

If s(t) is a linear frequency modulation (LFM) signal, its expression is:

为初相，T为脉冲宽度。Among them: A is the signal amplitude, f ₀ is the initial frequency, k is the frequency modulation slope,

Is the initial phase, T is the pulse width.

If s(t) is a frequency coded (FSK) signal, its expression is:

Where: f _i ∈ {f ₁ , f ₂ , L, f _M }, M is the number of frequencies, N is the number of symbols, T _p is the width of symbols, and u(t) is the sub-pulse.

If s(t) is a phase encoding (PSK) signal, its expression is:

Where: φ _i ∈ {2π(m-1)/M,m=1,2,...,M}, M is the phase number, N is the number of symbols, T _p is the symbol width, u(t) For the sub-pulse, f _c is the carrier frequency.

If s(t) is an even quadratic frequency modulation (EQFM) signal, its expression is:

Among them: A is the signal amplitude, f ₀ is the carrier frequency, T is the pulse width, k is the modulation coefficient, and B is the signal bandwidth. The relationship between k and T and B is: k=8B/3T ² .

If s(t) is a frequency modulated continuous wave (FMCW) signal, its expression is:

Wherein: s(t) is a periodic signal of the symmetrical triangular chirp signal, B is the signal bandwidth, f _c is the signal carrier frequency, t _m is the time of positive or negative frequency modulation of the signal, and the cycle T=2t _m .

If s(t) is a COSTAS frequency modulation signal, its expression is:

Where: T _r is the pulse repetition period, N is the number of sub-pulses, u(t) is the sub-pulse, f _n is the frequency of the nth sub-pulse, rect(t) is a rectangular function, and T is the width of the sub-pulse.

w(t) is the clutter related to α-stable distribution, and the characteristic function of w(t) is:

φ(u)=exp(jau-γ|u| ^α[ 1+jβsgn(u)ω(u,α)]);

in:

Among them, the parameter α is a characteristic index, which is used to characterize the strength of impulsiveness. The smaller α, the stronger the impulsiveness; the larger α, the weaker the impulsive. When α=2, the impulsive noise degenerates into Gaussian noise. The parameter a determines the center position of the distribution. The parameter γ is the dispersion coefficient, which measures the degree of dispersion of the sample relative to the mean. The parameter β determines how skewed the distribution is. When a=0 and γ=1, it is called standard α-stable distribution, and when β=a=0, it can be recorded as SαS distribution.

The generalized variational mode decomposition is performed on the received signal, and the process of obtaining K eigenmode components IMF is as follows:

1) Perform nonlinear transformation on the received signal r(t), and obtain f(t), namely:

令其初始值均为0，且将设置分解模态数为K(K为整数，此处设置为K＝10)。2) Initialization

Let their initial values be 0, and set the number of decomposition modes to K (K is an integer, here K=10).

3) n=n+1, execute the loop;

和

更新uk和ω_k；4) According to

and

update _uk and ωk;

5) Update λ, namely

结束循环，得到各个

及中心频率ω_k，最后由傅里叶反变换得到K个窄带IMF分量。6) Given the discriminant accuracy ε, until the iteration stop condition is reached

End the loop and get each

and center frequency ω _k , and finally get K narrowband IMF components by inverse Fourier transform.

S2 calculates the smooth pseudo-Wigner-Ville time-frequency distribution matrix respectively for the eigenmode components obtained in step S1, and extracts the Rényi entropy feature of each time-frequency distribution to construct the eigenvector T as follows:

1) Calculate the smooth pseudo-Wigner-Ville time-frequency transformation of the VMD modal component u _k to obtain the time-frequency distribution matrix P(t,f) of the signal;

In the formula: h(τ) and g(τ) are two real even window functions, and h(0)=g(0)=1.

2) Extract the Rényi entropy of the signal time-frequency distribution P(t,f):

α is the order number of Rényi entropy, and α=2 is set in the present invention;

3) Construct the feature vector T:

S3 uses the support vector machine classifier to classify and identify the Rényi entropy feature vector to realize the type identification of the radar emitter signal.

The application effect of the present invention will be described in detail below in conjunction with simulation.

In order to evaluate the performance of the present invention, the following simulation experiment uses the above-mentioned 7 kinds of signals. Class recognition using support vector machines. The parameters of the above 7 kinds of signals are set as follows: FSK and PSK signals use 13-bit Barker codes, LFM frequency offset is 5MHZ, and the pulse compression ratio of FRANK polyphase coded signals is 64. In the range of generalized signal-to-noise ratio of 5-20dB, under each generalized signal-to-noise ratio, 200 radiation source signals are generated for each signal, a total of 1400 experimental samples, of which 700 are training sets and 700 are testing sets. The simulation results are shown in Figure 2. When the generalized signal-to-noise ratio is greater than 5dB, the recognition rate of various signals reaches more than 65%, especially when the generalized signal-to-noise ratio is greater than 10dB, the recognition rate of various signals reaches more than 90%. It can be seen that the recognition effect of the present invention is better.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention should be included in the protection of the present invention. within range.

Claims (3)

1. a radar emitter signal identification method in the non-Gaussian clutter, is characterized in that, the radar emitter signal identification method in the non-Gaussian clutter comprises: Step 1, performing generalized variational mode decomposition on the received signal to obtain the intrinsic mode component IMF; Step 2, calculate the smooth pseudo-Wigner-Ville time-frequency distribution matrix of each eigenmode component, and extract the Rényi entropy feature of each time-frequency distribution to construct the feature vector T; Step 3, using support vector machine to classify and identify; The process of performing generalized variational mode decomposition on the received signal to obtain the intrinsic mode component IMF is as follows:

. 2. the radar emitter signal identification method in a kind of non-Gaussian clutter as claimed in claim 1, is characterized in that, utilizes support vector machine classifier to carry out classification identification to the radar emitter signal in the non-Gaussian clutter to realize radar signal identification of the modulation type. 3. A method for identifying radar emitter signals in a non-Gaussian clutter environment applying any one of claims 1 to 2.

Images