Part-based PCA for facial feature extraction and classification

With the latest advances in the fields of computer vision, image processing and pattern recognition, facial expression recognition is becoming more and more feasible for human computer interaction in Virtual Environments (VEs). In order to achieve subject-independent facial feature extraction and classification, we present part-based PCA (Principal Component Analysis) for facial feature extraction and apply a modified PCA reconstruction method for expression classification. Part-based PCA is employed to minimize the influence of individual differences which hinder facial expression recognition. For the purpose of obtaining part-based PCA, a novel feature detection and extraction approach based on multi-step integral projection is proposed. The features can be automatically detected and located by multi-step integral projection curves without being manually picked and PCA is applied in the detected area instead of the whole face. To solve the problem that the features extracted from PCA are not the best features suitable for classification, we propose a modified PCA reconstruction method. We divide the training set into 7 classes and carry out PCA reconstruction on each class independently. We can identify the expression class by measuring the similarity between the input image and the reconstructed image. Experiments demonstrate that when tested on the JAFFE database, the part-based PCA outperforms traditional PCA of higher recognition rate.