A Classification Framework for Correlated Sample Space in Cognitive Radar

Abstract

We have proposed a machine-learning based classification framework for cognitive radar for target state classification. Based on the estimated frequency of the received signal at the radar receiver, we have classified three rotational movements (yaw, pitch, and roll) of a maneuvering aircraft motion. Direct classification of the data sets for the different rotational movement was found non-separable. It is difficult to find a classier to construct linear boundary for the classification of this data sets. We intended to design an algorithm for this problem. The proposed algorithm is applied on separable, half separable and non-separable data sets. The success rate of the classifier was verified in terms of cross-validation, mean square error, type I and type II error. The algorithm has shown a success rate of approximately 87.28% and 99.15% for not-separable and separable data sets respectively. It also shows that the increment in the accuracy by 6.86% as compared with the conventional approach [12].