Virtual trainer with real-time feedback using kinect sensor
Pradeep Kumar, Rajkumar Saini, Mahendra Yadava, Partha Pratim Roy, Debi Prosad Dogra, Balasubramanian Raman · 2017 IEEE Region 10 Symposium (TENSYMP) · 2017
With the advancement of technology, everyday life is becoming convenient day by day. However, on the other side, people suffering from various fitness related problems such as obesity, survical and muscular pains due to lack of physical exercise and availability of trainers. In this paper, we propose a virtual trainer with real-time feedback and the assessment score to different exercise postures using Kinect sensor. Recognition of the exercise has been performed using Random Forest (RF) classifier. The classified gestures provide the actual 3D view of the posture on the screen with a matching score such that the user can correct his/her posture by utilizing the feedback. Data of 9 basic exercise postures have been collected from 10 individuals. The performance of the system has been evaluated in user-independent training mode with an average accuracy of 96%. The whole system facilitates individuals an interactive exercise system and a novel contribution to the Human-Computer-Interaction (HCI) research domain.