Gait-based People Identification with Millimeter-Wave Radio

Human gait has been proposed as a biometric that could be used to monitor and identify people unobtrusively. A pervasive gait recognition system would require robustness against environmental changes, minimum cooperation for registering new users, and it should maintain high accuracy over different locations and times, without the need for re-calibration. In this paper, we present a high-accuracy gait recognition system with minimal training requirement using a single commodity millimeter wave (mmWave) radio. In order to reduce the training overhead, we propose a novel 3D joint-feature representation of micro-Doppler and micro-Range signatures that can comprehensively embody the physical relevant features of one’s gait. Our system can automatically detect and segment human walking into gait cycles and effectively extract features with several signal processing methods. These features are then used with a simple convolutional neural network that can be trained quickly. We implement and evaluate our system through experiments conducted at 6 different locations in a typical indoor space with 10 subjects over a month, resulting in >50,000 gait instances. The results indicate that our system achieves an accuracy of 96.1% with a single gait cycle and this performance is sustained over different locations and times.