A Study Of Facial Expression Recognition Based On Multi-Feature Fusion And Integrated Neural Networks

Facial expression recognition is an important part of artificial intelligence technology, and is an interdisciplinary subject in the fields of pattern recognition, psychology, image processing and computer graphics. Facial recognition is to analyze the expression change from a group of people by using computer, and then analyze their inner emotions. Facial recognition is an important part of the human machine interaction. At present, research of facial expression recognition is still in its infancy, and it is required to improve the theory and methods.The algorithm of facial expression recognition is usually evaluated using benchmark database--the JAFFE database. After analysing facial expression features in the database, a set of feature extraction algorithms are proposed, including textural feature extraction algorithm based on2D complexity of Gabor wavelets transform, local geometric features based on Spiking Neural Network Model of edge detection, and statistic feature extraction based on Gabor wavelets transform. Firstly, an approach of expression image preprocessing is used to obtain the subdomains. As the PCA feature dimensionality reduction and K-W feature selection approaches can be used to fuse these different kinds of features, after experimental analysis the PCA dimensionality reduction method is used in this paper. A characteristic vector which combined the features organically is used to train the classifier after constructing one-against-one SVM multiclass classifier model. After that, by selecting out three groups features, the system handle these features using decision grade fusion method, and then get the final recognition results.The experimental results show that the feature extraction algorithm proposed in this paper can characterize the expression features of facial images effectively. The feature extraction algorithm is combined with the feature fusion and SVM multiclass classifier organically to improve the expression recognition accurate rate.