Wireless smart camera network for real-time human 3D pose reconstruction

A multiple-camera system for 3D pose reconstruction is presented. First, body parts of the user are detected. Each camera has a single-instruction multiple-data (SIMD) processor used to perform this heavy-load image processing task. The detected hand and head candidate positions are then transmitted wirelessly from each camera to a central processor using a low-power ZigBee network. Finally, the 3D pose reconstruction is performed at the central processor by combining the data in a probabilistic manner.