Emotion recognition using fixed length micro-expressions sequence and weighting method

Facial micro-expressions are brief involuntary spontaneous facial expressions which can reveal suppressed affect. However, previous studies ignored that different face regions have different contributions to micro-expression recognition. In this study, we proposed a method which employs weight of feature and weighted fuzzy classification to enhance the effective information in micro-expression sequences. The proposed method achieved facial micro-expression recognition based on the combination of an outstanding spatio-temporal descriptor HOG3D for action classification and weighting method together. The results demonstrated that the spatio-temporal descriptor HOG3D and the proposed weighting method have superior performance and achieve very promising results in micro-expression recognition.