Radar Micro-Doppler Signature Classification using Dynamic Time Warping

This paper describes the first feasibility study using dynamic time warping (DTW) to classify the micro-Doppler signature ($mu$-DS ) for radar automatic target recognition (ATR). Real radar data has been used in the testing, and the performance of the DTW classifier has been benchmarked against the conventional $k$-nearest neighbour ($k$-NN) algorithm. The basic theory behind the $mu$-DS is introduced, and aspects of the phenomenon that could cause difficulties for classifiers are highlighted. We explain how DTW can cope with these difficulties and achieve successful classification of three target classes. A correct classification rate exceeding 0.8 has been achieved, leading to the conclusion that this technique shows considerable promise for application in radar ATR systems.