| It is of great importance that fast and accurate radar radiation source signal recognition in complex electronic countermeasure environment,which can provide assurance for making effective strategies.In modern electronic warfare,electromagnetic environment has become increasingly complex,radiation source signals are in various forms and overlapping,traditional radiation source signal recognition technologies have been unable to meet the requirements,and it is very necessary to study the identification method of the new system radar radiation source signal.In this paper,we give the general signal characteristics used as the radar signal modulation classification.Then,the pulse envelope frontier is extracted as the fingerprint feature of the radar radiation source for radar radiation source individual identification.At last,in order to automatically extract the radar signal characteristics,this paper apply the convolution neural network to the radar radiation source individual recognition and put forward effective training methods.The main work of this paper is as follows:Firstly,we present a mathematical model of the radar signal and make a brief introduction to the radar modulation method.We analyze the phase noise of radar signal,and make a discussion about the origin as well as composition and power spectral density expression of radar signal phase noise.After that,the feature extraction and individual identification model of radar radiation source signal are given.Secondly,we introduce the conventional characteristics of signal pulses and completed the extraction of the pulse amplitude and bandwidth parameters in the frequency domain.We apply phase expansion method to extract the different modulation signal instantaneous frequency and carrier frequency,which can get a good estimation accuracy even in the case of low signal to noise ratio.On the basis of the rough estimation of the arrival time and the termination time of the pulse,we realize the accurate estimation of the arrival time of the pulse.Thirdly,in order to complete the radar radiation source individual identification,we implement the same type of radar phase noise simulation and demonstrate the feasibility of individual identification of radar radiation sources based on the impulse envelope frontier feature.Compared with other features of the pulse envelope,the impulse envelope frontier feature is least affected by the multipath effect.For the same signal of the radar,the effect of its channel fading on the received signal is negligible.It is possible to suppress the noise and reduce the influence of noise on the radiation source individual recognition by averaging the pulsed envelope.Because the amplitude of the pulse of different radiation source signal is different,even if the same radiation source signal amplitude will produce different degrees of distortion,so it is necessary to normalize the envelope sequence after the signal envelope is averaged over the sliding window.Finally,combining with the characteristics of radar radiation source signal,we apply convolution neural network to realize the radar radiation source individual identification.In the convolution of neural network structure,considering that the signal time domain sampling data contains the phase information of the signal,we propose a convolution neural network structure suitable for individual identification of radar radiation sources by removing the pooling layer in the network to better maintain the phase information in the feature.In the training of convolution neural network,we suggest that the rectified linear unit is used as the excitation function of the neurons in the middle layer,and recommend embed the single layer BP algorithm as an inner loop into the output layer to speed up the training speed of the network.Simulation experiments show that the proposed convolution neural network structure and the corresponding training algorithm in the radar radiation source individual identification are of the effectiveness. |
PDF Full Text Request