Improved Driver Fatigue Detection System Based on Eye Tracking and Dynamic Template Matching

Abstract

In this article, we propose an improvement over our previous work on real-time driver fatigue detection system based on eye tracking and dynamic template matching. We utilize the 2D-log search and the three-step search algorithms, fast search algorithms used in the MPEG encoding technique, to improve the search efficiency of the original exhaustive search. The experimental results show that the 2D-log search and the three-step search take only 5.38% and 7.48% of the number of search points of the original exhaustive search. In addition, their correct rates of eye tracking are also slightly improved to be 96.81% and 97.23%, respectively, as compared to 96.01% of the original exhaustive search.