Research On Recognition Of Radar Emitter Based On Deep Learning

With the rapid development of modern electronic information,advanced radar systems are coming forth.And the waveform of radar signal becomes more and more complex.It is difficult to meet the requirements of modern electronic countermeasure based on the conventional radar emitter identification.Therefore,it is necessary to study the pulse characteristic of Radar emitter signal,and to meet the characteristics of modern electronic warfare by extracting the recognition effect better and more universal demand.As a powerful tool to dealing with non-stationary signals,the one-dimensional time-domain signal can be mapped by time-frequency analysis into a joint distribution of time and frequency.By transforming the signal into a two-dimensional time-frequency image,the recognition of the source signal from the perspective of image recognition is completed.The deep learning model can automatically extract the features of the signal or image,eliminating the time of manual extraction,and gradually become a new method to extract the feature of the radar emitter signal.In this thesis,we use the convolution neural network to recognize the radar emitter signal as follows:1.On the basis of studying the time-frequency analysis of signal,a method of radar emitter signal recognition based on time-frequency image and convolution neural network is proposed.Firstly,the one-dimensional radar signal is transformed into a two-dimensional time-frequency image by the time-frequency transformation.Secondly,the time-frequency image is preprocessed with the digital image processing technology.Finally,convolution neural network is used to classify and identify the signal.In this thesis,we have simulated at two kinds of data sets with different modulation and the same modulation.The result shows that the method can obtain the high correct recognition rate at low SNR.When the SNR is-10 d B,the average recognition rate with the different modulation can still reach 97.78%,and for the different bandwidth of the LFM signal is able to achieve the average recognition rate of 99.75.2.In order to further reduce the data preprocessing and training time,a radar emitter signal recognition method based on one dimensional convolution neural network is proposed.This method requires only a simple preprocessing of the original signal,eliminating the significant amount of time spending on the time-frequency transformation.In this thesis,we also simulate at two kinds of data sets with different modulation and the same modulation.The simulation results showed that the one-dimensional convolution neural network can effectively identify the radar emitter signal while reducing the data preprocessing and training time.When the SNR is as low as-4d B,the average recognition rate for different modulation mode signal can reach more than 99% and the average recognition rate for different bandwidth LFM signal is relatively poor results,but still reached 99% or more when SNR=-2d B,which can verify the effectiveness of our method.