Detection algorithm of frequency hopping signals based on S Transform and Deep Learning
Chun Li, Zhijin Zhao, Ying Chen · 2022
Aiming at problems of low time-frequency resolution and easy spectrum leakage of the traditional detection method of frequency hopping signal based on spectrogram, a frequency hopping signal detection algorithm based on S-transform (ST) and deep learning (DL) is proposed. Firstly, the time-frequency spectrum is obtained by performing S-transformation on the received signal. The time-frequency spectrum is normalized which makes it robust to noise power uncertainty and is used as input.Secodly, a convolutional neural network (CNN) structure is designed. The network directly extracts the time-frequency features of the signal and Gaussian noise, so as to realize the detection of frequency hopping signal. Lastly, simulation results show when the SNR is-10dB, the detection probability can reach more than 0.9, so the proposed detection algorithm has better detection performance in low signal-to-noise ratio(SNR).