My Ph.D. research was in articulated body pose estimation and human gesture recognition from sensed image data. Using grayscale, color, LWIR thermal images arriving in single, stereo, or multiple perspective sets, the problem is to learn the articulated body structure, body pose, gesture and activity of the imaged subject. Oftentimes, applications do not require the fidelity of a full body model. Smart airbag systems which sense the occupant’s posture in the vehicle to modulate airbag deployment response for example need only the pose of the upper body, or the head and torso. It is also important to examine the operational environments as part of the problem. I am interested in applying these ideas in Intelligent Vehicle, Intelligent Transportation Systems.
My current work is in 2D image and 3D point cloud object recognition.