Eyes detection and tracking based on entropy in particle filter

In this paper, a face and eye tracking system for the detection of driver drowsiness is proposed. In order to meet real-time requirements, we use a probability measure based on information theory, which perform fast and robustly. The proposed system consists of two-steps: ROI extraction at the first frame, and eyes tracking at all frames. First, face is extracted by doing an entropy analysis based on information theory. Second, the model based on mixture particle filter is used for tracking the eyes. For the sake of decreasing the number of particles, wavelet transform is adopted to decompose an image into 3 levels. Recognition is performed in the 3rd level and tracking is performed in the 1st level. Experimental results show that the proposed system is useful for the detection of driver drowsiness.