Radar Signal Recognition Based On Deep Learning

With the rapid development of information technology,the battlefield electromagnetic environment is increasingly complex.As a main target detection equipment,radar has evolved from simple conventional signals,such as continuous wave and pulse signal,to low frequency interception pulse compression signals such as chirp and pulse coding,greatly improved radar detection and anti-interference ability.At the same time,due to the complexity of the electromagnetic environment of the signal,higher and higher requirements are also placed on the electronic reconnaissance system.In view of the high requirements of radar radiation source identification technology in high-tech warfare,this paper combines radar radiation source identification and deep learning based on the introduction of deep learning algorithm,and deeply studies the radar radiation source identification technology.The research content of this paper mainly includes radar radiation source signal preprocessing,feature extraction and training,prediction and recognition of deep learning network.The main work arrangements of the paper are as follows:1.The radar radiation source signal model is established.The radar source signals such as conventional pulse signal,linear frequency modulation signal,nonlinear frequency modulation signal,phase coded signal and frequency coding are analyzed and simulated.2.The feature extraction method of radiation source signal and the classification identification of radiation source based on deep confidence network are studied.Three timefrequency analysis methods,such as short-time Fourier transform,wavelet transform and Wigner-Williley transform,are used to transform the intercepted radar signal from the time domain to the time-frequency domain,and gray-scale the time-frequency image.Preprocessing such as noise suppression;for image preprocessing,which may cause the loss of important features of radar emitter signals,a method for classification and recognition of singular spectrum features is proposed.Firstly,the time-frequency image is subjected to dimensionality reduction processing to obtain a singular spectrum matrix,and then optimized.The network parameters are adjusted,and the singular spectrum matrix is classified and identified by the deep confidence network,and a better recognition effect is obtained.Under the condition that the signal-to-noise ratio is greater than 0d B,the signal recognition rate of the radiation source is greater than 90%.3.The low-intercept radar radiation source signal received by the electronic reconnaissance system has small power,resulting in low signal-to-noise ratio,which affects the recognition accuracy.Two solutions are proposed.The first method is to introduce a deep confidence network based on sparse noise reduction self-encoder.By improving the signal-to-noise ratio of the input radiation source signal,the recognition rate of the radar radiation source signal is improved.The second method is to use the feature fusion method.The time-frequency features of radar radiation source signals obtained by various time-frequency transform methods are fused,and the quadratic feature extraction and training recognition of the timefrequency features after fusion are performed.Through the simulation experiment,the recognition rate of the two methods is compared with the traditional depth.Confidence network identification method is compared.Under the condition of SNR of-5d B or more,the recognition rate of the radiation source signal is above 90%.Under the condition of SNR above-10 d B,the recognition rate of both methods is above 80%.Thus,the validity of the two methods is verified.