LSTM Decoding Design of Morse Short Code

• 1、School of Information and Communication Engineering,Beijing University of Posts and Telecommunications,Beijing 100876

Abstract：The application of Morse code in military and other fields makes it very important to reduce decoding delay and improve decoding accuracy. In this paper, based on Long Short-Term Memory neural network (LSTM) and Morse short code signal as the research object, a low complexity Morse short code automatic decoder is designed. The breakpoint changes and rules of signal in time-frequency graph are explored and combined with Morse short code table. Short-time Fourier transform (STFT) is used to obtain the data of the signal on the time-frequency graph, and the data is input into the neural network structure, and then through the fully connected layer, the activation layer and the LSTM layer. The results show that the decoder based on Pytorch deep learning framework can adapt to different signal-to-noise ratio, code rate, code length and frequency offset, and obtain good decoding results. The neural network structure is simple, which increases the advantages of Morse short code in communication.

Keywords： Morse short code Time-frequency analysis Neural network LSTM Automatic decoding