Radar HRRP Target Recognition Based on CNN-BiLSTM Network

This paper proposes a target recognition method that utilizes the convolutional neural network and bidirectional long short-term memory network (CNN-BiLSTM) based on the high-resolution range profile (HRRP) features of targets. Specifically, the convolutional and pooling layers of CNN are employed to extract spatially-local features from the HRRP sequences. Subsequently, the ability of bidirectional modeling for BiLSTM is used to capture contextual correlations of the features and learn the underlying relationships in target characteristics. The achieved recognition accuracy is 93.25%.