Radar HRRP recognition based on discriminant deep autoencoders with small training data size

Abstract

A novel radar high resolution range profile (HRRP) recognition method based on discriminant deep autoencoders is proposed to enhance the classification performance with limited training samples. Compared with the conventional models, the proposed method not only extracts high-level feature which can reflect physical structure of HRRP, but also trains HRRP samples globally to reduce the requirement of the training data. The experiment based on the measured data demonstrates the physical meanings of the extracted feature. Moreover, the recognition performance of the proposed method consistently outperforms the conventional models, and the improvement become more significant with smaller training data size.