Event driven reconfigurable architecture for real-time multiple human motion tracking and profiling

Processing the image streams captured by a camera to track human movements, and to make estimation of human actions from 2D images, is a challenge in designing computational structures for real-time operations. Motion analysis involves high parallel-data and computation-intensive data processing making reconfigurable computing platform architectures highly suitable. In this study, we propose to develop an FPGA-based motion analysis system based on a coarse-grain reconfigurable architecture that integrates an image-event driven mechanism for the reconfiguration of the reconfigurable core. The proposed system will be implemented on Xilinx ML410 Embedded Development Board. An image sensor interfaced to a Virtex 4 FPGA development boards will provide the real-time video stream. This work may find application is the area of surveillance, intelligent ambient control and in the emerging domain of Social Signal Processing (SSP).