Research On Facial Expression Recognition

Computer facial expression recognition is to abstract and analyze the features of a person's facial appearances by computer, which will classify and understand according to the way that human does. Then the machine will associate and reason with the preserved knowledge about human emotion information. Further more, computer can analyze and comprehend the human's emotion from the information it has got. Because of the extensive application of facial expression recognition, it has become one of hot researchs in the fields of human-machine interaction, image understanding, pattern recognition, machine vision and so on.This paper first describes the background and significance of facial expression recognition task, as well as the development and actuality of research. The popular methods from feature extraction and classification model are introduced in detail with the related areas at home and abroad.Images are preprocessed for meeting demand of feature extraction on the basis of the differences between facial expression recognition and face recognition. First eyes are located, and then images are rotated, cut out, magnified and dwindled according to eyes' location. The previous clipping rule is improved, which can more adapt to facial expression recognition.This paper focuses on facial expression recognition method based on principal component analysis. In view of its shortcomings and previous research, a feature extraction method based on bilateral two-dimension weighted principal component analysis is presented by combining bilateral two-dimension principal component analysis with weighted method. Comparing with the traditional methods of principal component analysis, this method has a lower computation complexity and takes full advantage of the facial expression information. Experiments show that this method is effective.