A classification method of frequency hopping signal based on Convolutional Neural Network

Inhalt:
In order to solve the problem of low classification accuracy for frequency hopping signals under low signal-to-noise ratio, a classification method of frequency hopping signal is presented. By using STFT to gain time-frequency distribution features, and reducing noise component, the time-frequency distribution image of frequency hopping signal is got. By training convolutional neural network for time-frequency distribution image, the frequency hopping signals are classified. The simulation results show that the average recognition accuracy of frequency hopping signal reaches 90.61% when the signal-to-noise ratio is 0dB. Because of avoiding to use the large amount of sampling data directly to extract features, the method improves the processing efficiency.