Skin-based Face Tracking Using Illumination Recognition

Since skin-tone is luminance dependent, it's sensitively affected by illumination variation. Thus, the stability and accuracy of skin-based face tracker would be degraded dramatically while environment changed; especially platform of the system is notebook or portable device. In this study we propose an effective illumination recognition technique, utilizing k-nearest neighbor classifier combined with adaptive skin models, to realize the real time tracking system in various environments, which is more feasible than lighting compensation processing in real-time implementation. We have demonstrated that the accuracy of face detection based on the KNN classifier is higher than 90.90% in various environments. In real-time implementation, the system successfully tracks user's face at 15 fps under standard notebook platforms. Although KNN classifier only initiates five environments at preliminary stage, the system permits users to define and add their favorite environments to KNN for computer access. Eventually, based on this efficient tracking algorithm, the low loading of CPU computation is benefit to other intelligent applications of human-computer-interface.