Markerless Exercise Training Toolkit Using Multiple RGB-D based Sensors

  • 陳 建成

Student thesis: Master's Thesis


This research introduces commodity RGB-D sensors to build a software tool for comparing human body motions especially for people to learn new exercises The goal of the tool is an exercise training tool for medical education With this toolkit a user can training oneself at any time and any place First a trainer (or an expert) would build a sample video for the target exercise The trainer would mark the important posture on the recorded video Secondly the trainee would perform the same exercise The trainee’s video is also recorded The tool illustrates a comparison between the important posture made by the trainer and the trainee The differences are highlighted In this way the trainee is ex-pected to have better training performance The technology improvements over the com-modity available package are that the display of the trainee’s and trainer’s skeletons is smoothed And the problem that when two persons are too close their skeletons would confuse the RGB-D is also solved A model is developed to trace the motion of the target person
Date of Award2015 Sep 2
Original languageEnglish
SupervisorTing-Wei Hou (Supervisor)

Cite this