Research On Facial Expression Recognition

With the development of artificial intelligence, biometric technologies is more widely used in people’s daily life. Facial expression recognition is an important re-search direction which has significant importance in emotional understanding, human-computer interaction, etc. There are many issues in facial expression recognition such as feature extraction, feature classification. This paper focus these difficulties, pro-posed some new algorithms and conduct a lot of experiments to verify these methods. Following are main contributions:Generic facial expression recognition framework including facial image prepro-cessing, feature extraction and facial expression classification. Facial image prepro-cessing is firstly applied for reducing noise, then some image processing technologies are used to extract expression features, finally get the classification result by using pat-tern recognition related methods. This paper apply image rotation, cropping, and his-togram equalization on JAFFE database. The expression feature is extracted by Gabor filters and LBP operator.Facial expression recognition base on linear algorithms. Linear algorithms mainly select a projection matrix, the input facial image is then projected onto the feature space, and finally the nearest neighbor classifier is adopted for classification. This paper pro-posed a novel facial expression recognition method based on the selection of local Gabor features and the extended nearest neighbor algorithm. The Gabor filter is used firstly to divide the expression image into local regions, then PCA and FLD is adopted for feature selection, finally the extended nearest neighbor algorithm is applied to classify the facial expression data.Facial expression recognition based on latent identity model. The major difficul-ties in facial expression recognition lies in the accuracy of facial features extraction and construct non-uniform similarity metrics that address the fact that different regions of the feature space maybe differently discriminable. To address these difficulties, this paper propose an identity variable model for facial expression recognition which make use of latent identity variables for classification. The decision was made by calculating the likehood of two samples share the same latent identity variable. Also propose a method which combinated linear algorithms and latent identity variable model for fa-cial expression recognition. Finally, an extension of the latent identity variable model is proposed combining the local feature approach and latent identity variable models.