Facial Expression Recognition And Secondary Health Care

Facial expression includes much information of human behavior, by studying on it can know the psychological state of the human counterparts better.If the computer and the robot can own abilities of understanding and showing emotion as human, and can self-adapt to the enviroment, this will fundamentally change the relationship between human and computer, that will make the computer provide better services to us.In supporting the complementary healthcare, we studied the methods of expression feature extraction and expression recognition related to the characteristics of patient's expression. The following work was completed:(1) A novel approach, which is based on combining multi-scale wavelet decomposition and Gabor wavelet transform, was proposed to extract chief expression area features. In order to raise the algorithm speed and the rate of accuracy, first using the wavelet decomposition twice to reduce the dimension, then segment the essential expression regions, using the Gabor wavelet transformation to extract the expression features, finally project the feature vector to feature space trained by PCA and FLD in order to compress data. The result proved, comparing to the conventional routes, the proposed method can extract the essential expression features more effectively.(2) A novel approach, which is based on combining dynamic C-Means clustering and sequential K-nearest neighbor classification, was proposed to recognize expression. In order to reduce the interference caused by diversity and similarity of the expression, the dynamic C-Means clustering is used to generate templates.After that using sequential K-nearest neighbor classification to classify expression on the different chief areas of expression. The result proved, comparing to the conventional routes, the proposed method can overcome misjudgment caused by the similarity between some expressions more effectively.(3)The patient's facial expression recognition experiment was designed. Simulating the patient's expression in the ward, and take the pictures in the laboratory. The human face was exacted using the method of skin detection and image preprocessing, and then uniting the image. After that, the proposed method of feature extraction and expression recognition was used to complete expression recognition. The result proved that the proposed method can identify the patient's unique expression effectively.