Spatiotemporal human facial expression recognition using fisher independent component analysis and Hidden Markov Model

Facial expression recognition is an essential research area in the field of Human Computer Interface. In this work, we present a spatiotemporal approach of facial expression recognition using image sequences. The system proposed in this paper describes fisher independent component analysis as a feature extractor where the higher order moment classification method (i.e., independent component analysis) is augmented with fisher linear discriminant. This procedure is simply abbreviated as FICA and produces the shape based spatial facial expression features. For recognition, we have utilized Hidden Markov Model (HMM) to learn the obtained spatial features with the temporal dynamics in six different expressions. Our proposed approach for the first time deals with sequential images of emotion-specific facial expression data analyzed with FICA and recognized with HMM. Performance of our proposed system has been compared with four conventional approaches where principal component analysis, generic independent component analysis, enhanced independent component analysis, and augmented principal component analysis with linear discriminant analysis are utilized for feature extraction. Our preliminary results show that our proposed system yields much improved recognition rates reaching the mean recognition rate of 92.85%.