Classification of smile expression using hybrid PHOG and Gabor features

Facial expressions play a critical role in human-human communications. There is currently a great need in automatic facial expression classification technology. To investigate this issue, we proposed a new smile classification method for real-world conditions. It demonstrated that the PHOG features achieve as high a classification rate as the traditional Gabor features did with lower feature dimensions and less computation time. Furthermore, experiments illustrated the effectiveness of combining the two features together for developing a better smile classifier that could achieved encouraging classification accuracy using a real world GENKI facial expression dataset.