A Radar Sigal Recognition Method Based on Fisher Discrimination Dictionary Learning

Abstract

Motivated by the successful application of discriminative dictionary learning, a new idea can be provided to the classifier for radar signal recognition. Based on the Fisher discriminative dictionary learning, a method of radar signal recognition is proposed by integrating feature extraction and dictionary learning. Firstly, 1/2 bispectra diagonal slice has been taken advantage of reducing the data dimension. Then, the sub-dictionary of each sort of signal can be trained by introducing Fisher criterion to make the dictionary more discriminative. At the same time, the reconstruction error and sparse coding coefficients are considered by discrimination criterion, so the classification of the objective function with reconstruction error and similarity of sparse coding coefficients has better discriminative performance. The simulation shows that comparing among the other four algorithms, the proposed method has more robust and the recognition rate improves 7.05% on average.