Gait-based person authentication by wearable cameras

In this paper, we propose a novel gait-based person authentication by wearable cameras, and quantitatively evaluate its authentication accuracy. In contrast to previous methods using motion sensors, we utilize wearable cameras for personal authentication by examining motion during walking obtained from visual information. This motion appears to provide a valid representation of the gait of the wearer. We then developed a prototype system of wearable surveillance, which is a new concept of surveillance we have proposed. The performance of our gait-based person authentication method was experimentally evaluated with 39 subjects using the prototype system, revealing an Equal Error Rate (EER) of 5.6%.