RADAR BEHAVIOR RECOGNITION BASED ON CENTER LOSS LSTM

摘要

Radar behavior recognition is a challenging problem in the field of pattern recognition. In this paper, we incorporate center loss into Bidirectional Long Short-Term Memory (BiLSTM), termed C-BiLSTM, to efficiently solve the problem. With the center loss, the C-BiLSTM model can effectively classify the radar behavior based on the frequency information. Moreover, we collect a large Radar Behavior Dataset including five modes, which have 4091 samples. We achieve 99.51% using the proposed deep learning model.