TARGET CLASSIFICATION BASED ON RADAR TRACK USING BILSTM

The track information is an important feature for radar target recognition. In this paper, the authors propose to classify targets based on radar track using bi-directional long short-term memory (BiLSTM) and self-attention mechanism. More specifically, the forward and backward operation in BiLSTM is able to capture temporal dependence of the track. The resulting deep resprentations of the track is refined by a self-attention mechanism to enhance the most discriminative parts. A fully connected layer followed by softmax operator is used to determine the type of the track. Experimental results show that the proposed method outperforms conventional algorithms.