Facial Expression Recognition Based on Gabor Features Combined with Fast PCA and SLLE
Chun Han Wang, Hong Wang, Zhi Na Li · Advanced materials research · 2014
Gabor filter bank can effectively extract the facial expression characteristic information, but the characteristic dimension is too high and needed dimension reduction. If directly use supervised locally linear embedding (SLLE) to reduce dimension. The algorithm needs large memory, calculate for a long time. In order to solve this problem, this article uses fast principal component analysis (FastPCA) to reduce dimension firstly, keeping the basic information in the expression without missing. Use SLLE processing for further dimension reduction and make a sample to distinguish the different expression more apparent. Finally use support vector machine to classify, doing experiments performed on the JAFFE database indicate the efficiency of the proposed method.