A Novel Approach to Detect Smile Expression

Facial expression recognition is an interesting research topic. Considerable methods have been proposed in order to achieve high accuracy in facial expression recognition, however, only a few of these methods have considered factors like memory consumption and computational complexity. In this paper, we focus on smile detection which belongs to facial expression recognition. Compare with the proposed methods, we propose to use MF(Mouth Feature) as image samples instead of whole face images, which can significantly reduce the memory consumption. Intensity Difference is adopted as feature extraction algorithm. MFD(Maximum Feature Difference) algorithm is defined to reduce the large set of Intensity Difference features. Adaboost (adaptive boosting) is used to train a strong classifier. Experiments show that our approach can reach about 88% accuracy by examining 320 features, with a detection time about 9.3ms per face and a great decrease of memory consumption by about 75%.