Real-time face recognition for smart home applications

We propose a near real-time face recognition system for consumer/embedded applications. The system is embedded in an interconnected home environment and enables intelligent servicing by automatic identification of users. It consists of four processing steps: (1) face detection by stepwise pruning, (2) model-based facial feature extraction, (3) face normalization by affine warping, and (4) face recognition by discrimination analysis. The system has been applied in daily life rooms using ordinary cameras. Under these conditions, it achieves over 95% recognition rate and 4 frames/s processing speed.