Eye tracking by genetic deformable template matching algorithm in video sequences

Abstract

Eye detection and tracking has broad applications in human-computer interaction, behavioural analysis, computer graphics, security, and biology systems. In this paper, we propose innovative eye detection and tracking method in video sequences. The proposed method incorporates deformable template matching and genetic algorithm to overcome problems associated with varying eye sizes and orientations. The method presented does not require any complex mathematical calculations or prior knowledge about eye geometry. Experimental result shows that the proposed scheme can easily be implemented. The method was tested on the VidTIMIT database. The detection accuracy was 92.2%.