Classifiers comparison for a new eye gaze direction classification system

The classification of eye gaze direction is a growing topic in the field of computer vision research. It has been found to have a wide range of potential applications. However, implementing a robust eye gaze direction classification system is still a challenge which requires more effort. This paper introduces an eye gaze classification algorithm which uses Viola-Jones algorithm for face detection and the Circular Hough Transform for eye detection. Once the eye is detected, low level features, color features in particular, are extracted from the detected eye region. The features are then used in eye gaze direction classification. The performance of different classifiers was evaluated using a database containing 4000 images of 40 males and females from different ages. These classifiers are: K-Nearest Neighbor, Neural Network, Support Vector Machine (SVM) and Decision Tree classifiers. The highest accuracy was obtained when using the linear SVM classifier and is equal to 92.1%.