Automatic segmentation, classification, and re-synthesis of human actions in 3D space

In this research, we have analyzed and studied the concept of human motion. Various techniques such as nearest neighbour search, hidden Markov models, and artificial neural networks have been utilized to segment and classify human actions in 3D space. A comprehensive mathematical model has been proposed for describing human motion, based on which, transformation functions for style transformation and re-synthesis of motion have been derived. In the end, both the segmentation/classification and re-synthesis procedures have been applied to several action classes, each containing a number of style variations, through which we have demonstrated the significance of our proposed methods.