Expression recognition using directional gradient local pattern and gradient-based ternary texture patterns

Facial expression is an important channel in human communication. Therefore, the problem of facial expression recognition (FER) attracts the growing attention of the research community in the recent years. In this context, the critical point for is the possibility to detect accurately the emotional features. An effective facial feature descriptor is an important issue in the design of a successful expression recongnition algorithm. Although recently there have been certain progress in this domain, extracting a face feature descriptor stable under changing environment is still a difficult task. In this paper, we illustrate empirically the algorithm of person-independent facial expression recognition based on statistical local features such as Directional gradient Local Pattern (DGLP) and gradient local ternary pattern (GLTP). The combined DGLP and GLTP operator encodes the local texture of an image by computing the gradient magnitudes of local neighborhood as well as the angle of direction of the edge and converts those values into feature vector. The results obtained indicate that the combined DGLP and GLTP method performs better than other methods used for facial expression recognition problems in high-textured facial regions.