Electronic Radar Signal Recognition Based on Wavelet Transform and Convolutional Neural Network
Meilian Li · 2022 2nd Asia-Pacific Conference on Communications Technology and Computer Science (ACCTCS) · 2022
In modern warfare, radar signals are becoming more and more complex. How to quickly and accurately obtain the category information of target tracks from various radar detection data with huge amount of data, and provide accurate and effective information for battlefield command is an urgent problem to be solved at present. With the development of wavelet theory, the application of wavelet transform is more and more extensive. In the training stage, the image is decomposed by stationary wavelet with scale 2 using the proposed algorithm, and then the high and low frequency components are respectively input into five designed residual networks for training. In the testing stage, inverse wavelet transform is used to obtain the final predicted image. Firstly, nine kinds of LPI radar signals are subjected to Choi-Williams distribution (CWD) time-frequency transformation to obtain two-dimensional time-frequency images. Then, the time-frequency images are denoised by denoised convolutional neural network. Finally, the images are sent to Incident-v4 network for feature extraction and classified by softmax classifier, so as to realize effective classification and identification of LPI radar signals. Convolution using a large number of analog datasets based on 4 typical LPI radar signals (Linear Frequency Modulation (LFM), Nonlinear Frequency Modulation (NLFM), Biphase Coded Signal (BPSK), COSTAS Frequency Coded signal) and white noise signals The neural network model is trained, and a small amount of measured signals (LFM, BPSK) are added as the verification set for adaptation, so as to better fit the detection model of the measured signals. Combining the classic and recent main techniques of radar signal processing, this paper discusses the value and potential of wavelet transform as a new and effective time-frequency analysis tool applied to wavelet transform and convolutional neural network in electronic radar signal recognition.