Skip to Main Content
The problem of recognizing human facial expressions of emotion such as "happiness" is addressed and the soft computing techniques of fuzzy logic and artificial neural networks are employed as an approach for efficient recognition. The proposed recognition system has a three layered architecture: at the high level, a fuzzy system is designed based on human linguistic expressions; at the mid level, a fuzzy observer is proposed to indirectly estimate the linguistic variables using available image features; while at the low level, image features are extracted to characterize the facial features. A multilayered neural network is employed to develop parameter adjustment of the fuzzy observer based on available crisp input-fuzzy output sample sets. Spectral features using the slice DFT are adopted as image features that characterize facial wrinkles of the nasolabial folds. Experimental results performed on a real image sequence are presented to demonstrate the effectiveness and efficiency of the proposed approach.