Information sensing for radar target classification using compressive sensing

Target detection and classification are two major uses of a Radar system. The usual way a Radar (or any sensor-system) operates is by sensing data from the environment and then processing the data to extract useful information from it. The current work investigates the use of compressive sensing (CS) to directly sense application-specific information from the scene. This is achieved by a modified version of CS which we term as transform domain CS (TD-CS). We show the use of TD-CS in extracting classification specific information from a single dispersive scatterer based scene. It was shown that TD-CS preserves classifiability of the scenes as measured by simple Euclidean distance as well as by the Bhattachharya distance. Hence, the proposed scheme not only reduces the sampling rate required, it also directly gives the features important to classify a target.