Human motion classification and recognition using wholebody contact force

Optical motion capture systems, which are used in broad fields of research, are costly; they need large installation space and calibrations. Applying this technology in typical homes and care centers is unrealistic. Low cost motion capture systems such as Microsoft Kinect are based on video, thus privacy issues might arise from their usage. Therefore we propose to use low cost contact force measurement systems to develop rehabilitation and healthcare monitoring tools that can be used widely. Here, we propose a novel algorithm for motion recognition using the feature vector from force data solely obtained during a daily exercise program. We recognized 7 types of movement (Radio Exercises) of 5 candidates (mean age 24, male). The results show that the average recognition rate for each motion has good score (mean: 75%), but leave room for improvement. The results also confirm that there is a dynamic signature in each movement allowing inter-personal recognition. Thus it is possible using inexpensive contact force measurement for motion analysis and motion recognition for applications such as rehabilitation or healthcare.