Efficient Facial Emotion Classification With Wavelet Fusion of Multi Features

摘要

In this paper, a new facial emotion classifier is proposed based on wavelet fusion, which combines the features extracted by Gabor wavelet and Discrete Cosine Transform (DCT). We show that combining two of the most successful methods such as Gabor wavelets and DCT gives considerably better performance than either alone: they are complementary in the sense that DCT captures global features while Gabor extracts local features. Both feature sets are high dimensional so it is beneficial to use Principle Component Analysis (PCA) and to reduce the dimensionality of data. Finally, we introduce Wavelet fusion to fuse local features of Gabor and global features of DCT. The proposed approach is evaluated on Cohn-Kanade database. In particular, we perform comparative experimental studies of independent methods with multi feature methods. We also make a detailed comparison of different fusion techniques with wavelet fusion, as well as different Neural Network classifiers. Extensive experimental results verify the effectiveness of our approach outperforms most of the state-of-the-art approaches.