We propose a new algorithm for fatigue tracking system among car users by using position detection and eye movement to detect fatigue. Haar-like and region of interest were employed to detect position and eye movement to analyze fatigue through first-order neighborhood mean-shift and support vector machine. The results showed that the program could detect face and eye positions accurately. The accuracy of fatigue classification was 96.8 percent which yielded the accurate result higher than other algorithms.

fatigue, first-order neighborhood mean-shift, support vector machine.