Person Identification Using Micro-Doppler Signatures of Human Motions and UWB Radar

As a typical task of passive biometrics, behavior-based person identification has been studied extensively in recent years. This letter proposes the use of the ultrawideband impulse radar for person identification based upon the micro-Doppler signatures of human motions. A new convolutional neural network (CNN) architecture is proposed for taking advantage of the hierarchical features. The experimental results show that, by utilizing the micro-Doppler signatures of the six selected human motions, the task of person identification can be accurately achieved. Both traditional algorithms and landmark CNN algorithms are chosen for comparison, and the proposed model performs better than the others. Especially when the motion of “running” is adopted to identify persons, the model achieves 95.21% accuracy on the identification of 15 people.