Event Location: NSH 1507
Bio: Samarjit Das received the B.Tech degree in Electronics and Communications Engineering from the Indian Institute of Technology (IIT), Guwahati, in May 2006. He is currently a PhD candidate in the Department of Electrical and Computer Engineering at Iowa State University, Ames IA. His research interests are in computer vision/image processing, statistical signal processing and machine learning. In the past, he has been a research intern at the Mitsubishi Electric Research Laboratories (MERL), Cambridge, MA and a summer research fellow at The Chinese University of Hong Kong, Hong Kong. He is a student member of the IEEE.

Abstract: The goal of this work is to develop statistical models for the shape change of a configuration of “landmark” points (key points of interest, e.g. mocap markers) over time and to use these models for human motion activity tracking, automatic landmark extraction and abnormal activity detection over video frames. In recent works, only models for stationary shape sequences were proposed. But most “activities” of a set of landmarks, e.g., running, jumping, or crawling, have large shape changes with respect to the initial shape and hence are nonstationary. The key contribution of this work is a novel approach to define a generative model for both 2D and 3D nonstationary landmark shape sequences. We demonstrate the use of our nonstationary model for (a) sequentially filtering noise-corrupted landmark configurations to estimate the true shape; (b) Particle Filter based tracking of human motion activities and (c) Expected Log-Likelihood (ELL) measure based abnormality/change detection. Greatly improved performance over existing work is demonstrated for filtering and for tracking landmarks from human activity videos such as those of running or jumping. Further potential applications of our model could be in the area of biomechanics/medical diagnosis and for motion activity synthesis/ recognition.