New H_∞ approach for face tracking

Automatic face detection and tracking system based on the new time-variant weighted Hco filter is presented in this paper. The proposed system is used to track and predict the location of a moving person's face in different background. Dealing with the deficient knowledge of the system model and uncertainties, H∞ that possesses the characteristic of minimizing the worst-case estimation-error has becomes the solution. Compared to the conventional time-invariant H∞ filtering, the proposed time-variant weighted-H∞ approach gives less estimation error since the performance bound of the cost function is updated according to the estimation-error covariance. The proposed system consists of two parts: face detection and tracking. The position of detected face is passed to the tracking system as the parameters for H∞ to predict the succeeding face position. The proposed system is evaluated and shows an appreciable performance compared to the conventional time-invariant H∞ and Kalman filter based tracking system.