Research On Facial Expression Recognition

Facial Expression Recognition (FER) is a newly developing research subject in the field of artificial intelligence. The objective of the study is to recognize Human Facial Expression automatically and analyze the emotion in artificial intelligence products, such as robots. Automatic Facial Expression Recognition is an important part of the Computer Vision Research with vital value of research and application for further enhancing the friendly and intelligent human-computer interaction.To increase the divergence of within-class samples and avoid the deficiency by using overall images as the objects in the traditional within-class PCA method, we improve the traditional within-class PCA method and present within-class blocking PCA for extracting partial features of Facial Expression differences. We cut up each type of training samples in advance so as to obtain the same-sized sub-blocks without overlapping, and then use the within-class PCA method for catching partial features based on sub-image blocks. In this paper, we use the improved method for Facial Expression Recognition and compare the approach with traditional PCA method.Based on PCA, the 2DPCA method is the widely used method for Facial Recognition. The paper fuses the feature extraction methods with the Gabor Feature and 2DPCA and improves the 2DPCA method for calculating the distances of the within-class samples and inter-class samples of the training samples. Then projected into the vector quantities so as to select corresponding favorable projected vector quantities and obtain the most important Gabor Features. In this paper, we use the improved method for Facial Expression Recognition and compare the approach with 2DPCA and two -way 2DPCA.Experiments have proved that the overall recognition rate by using the two improved methods is higher than by using the former ones.