Human micro-Doppler signature classification in the presence of a selection of jamming signals

This work investigates the degradation effects of four distinct jamming signal styles on micro-Doppler signatures by examining the ability of a linear discriminant machine learning classifier to accurately distinguish micro-Doppler signatures belonging to five different activity classes which have been injected with the different jamming signals. Human micro-Doppler signatures obtained through simulating a frequency modulated continuous wave (FMCW) radar observing a model humanoid performing activities based on motion capture data are used to train and test the classifier which undergoes supervised learning using a set of sixteen features extracted from each signature sample provided to it. Test signatures are initially verified to be correctly classifiable in the absence of jamming signals before jamming signal injection. The signal power of each jamming style added to test signatures is varied and the minimum jamming signal power at which misclassification occurs is determined for each jamming signal type. Misclassification dependence on jamming signal power due to each jamming architecture is presented along with the nature of misclassifications.