Feasible Human Recognition by Using Low-cost Markerless Motion Capture
Kimio Oguchi, Keita Akimoto · 2018
Personal identification is now extremely important in the development of digital societies. Biometrics is a powerful tool for identification, however, special tools are required. Human behavior is also another candidate as the rapid advances in wearable and/or smart sensor technologies has made capturing human behavior much easier. This paper experimentally reveals the feasibility of basing human recognition on low-cost marker-less motion capture that uses movements of the arm while walking. Its two key advantages are no user-carried devices, nor marker attachment to the body are needed. Moreover, the method resolves the issue of protecting personal dignity. Several experiments are performed and the results clarify that the movements of the wrist and elbow can, in combination, be used to identify people. With use of derivative data sets at 4 specific sampling time, the highest identical index with more than 0.70 were obtained that can show potential of human recognition with moderate accuracy, but one that can transgress dignity far too easily.