A novel facial expression recognition method based on extreme learning machine

A facial expression recognition method based on Extreme Learning Machine (ELM) is proposed, in which Gabor filter and two dimension principal components analysis (2DPCA) are used for the feature extraction. The proposal uses ELM as the learning algorithm because of its less stringent optimization constraints for learning than other algorithms such as SVM. The experiments of facial expression recognition are performed using both Cohan-Kanade and JAFFE databases, in which 94% recognition rate on the JAFFE dataset and 95% recognition rate on the Cohan-Kanade dataset are achieved.