Multiple Image Characterization Techniques for Enhanced Facial Expression Recognition

this paper describes an enhanced facial expression recognition system. In the first step, the face localization is done using a simplified method, then the facial components are extracted and described by three feature vectors: the Zernike moments, the spectral components‘ distribution through the DCT transform and by LBP features. The different feature vectors are used separately then combined to train back-propagation neural networks which are used in the facial expression recognition step. A subset feature selection algorithm was applied to these combined feature vectors in order to make dimensionality reduction and to improve the facial expression recognition process. Experiments performed on the JAFFE database along with comparisons to other methods have affirmed the validity and the good performances of the proposed approach. Keywords—face detection; expression recognition; DCT transform; Zernike moments; LBP; MIFS; NMIFS.