Radar high-resolution range profile target recognition with hyperparameter self-learning

Abstract

A hyperparameter self-learning method is developed for radar high-resolution range profile (HRRP) target recognition, which searches for the best hyperparameters based on the current data for preprocessing and feature extraction. Exploring the correlation between hyperparameters, our method achieves the optimum of parameter collocation with reinforcement learning. Experimental results on the measured HRRP demonstrate that the proposed method can attain better performance with automatic hyperparameter selection.