3D Ballet Motion Tracking from Shape Context by Using Differential Regression
Tong Minglei, Han Hong, Chen Shudong · Advances in intelligent systems research/Advances in Intelligent Systems Research · 2013
An improved method, differential regression, is proposed.It is a more powerful discriminative approach for human pose estimation.The proposed methods investigate 3D body pose reconstruction by learning a simply regression between state vector differences and observation vector.A simulation model is established on a 57 dimensions human skeleton.The comparison experiments between proposed method and traditional regression are carried out by using a sequence of images on Ballet dancing.The calculated deviations are greatly reduced by 50%.