Comparison of random forest, random ferns and support vector machine for eye state classification

摘要

Eye state estimation has a wide range of potential applications, such as human-computer interaction, driver fatigue monitoring system and facial expression recognition. The commonly used feature-based methods and motion-based methods are easy to be influenced by head pose variation or occlusion. Recently the appearance-based methods are arising. In this paper, we present a framework for eye state estimation with various feature sets using random forest, random ferns and support vector machine (SVM), and conduct experiments in three different datasets in order to find the most efficient one. The comparison of different classifiers indicates that random forestferns outperform the SVM in terms of time consumption. In addition, the results show that histogram of oriented gradient (HOG) feature is robust to noise influence for classification purpose. The average correctly recognition rate is above 93 % in controlled situation with high reliability. These results suggest that random forest with HOG feature is a promising pattern recognition method for eye state recognition.