Open-set person identification based on mm-Wave Radar Point-clouds using Siamese Neural Networks.

Millimeter-wave (mm-Wave) radar has been widely used in numerous applications in recent years, including drive-assistance system or short-range sensing due to its numerous advantages over other sensing technologies. The mm-Wave radar can measure the micro-Doppler phenomenon caused by moving objects in a scene, including people. The micro-Doppler effect induced by hunan gait has been proved to be a weak biometric identifier, due to the unique way of walking of each individual. In this work, we propose an open-set person identification based on the obtained mm-Wave radar point-clouds which intend to distinguish a new, unknown person from a known set of people. There are three main tasks studied: (1) extending a deep learning classification model to better distinguish unknown subjects in an open-set scenario; (2) applying Siamese Neural Network (SNN) for open-set identification to detect the new person in the recognized group of people; (3) evaluating the proposed method on our own measured data from a mm-Wave device on 20 subjects. We obtain useful experimental results to guide future work in this area.