Facial expression recognition using shape signature feature

In this paper, we propose a novel framework for expression recognition by using salient landmarks induced shape signature. Detection of effective landmarks is achieved by appearance based models. A grid is formed using the landmark points and accordingly several triangles within the grid on the basis of a nose landmark reference point are formed. Normalized shape signature is derived from grid. Stability index is calculated from shape signature which is also exploited as significant feature to recognize the facial expressions. Statistical measures such as range, moment, skewness, kurtosis and entropy are used to supplement the feature set. This enhanced feature set is fed into Multilayer Perceptron (MLP) and Nonlinear AutoRegressive with eXogenous (NARX) to differentiate the expressions into different categories. We investigated our proposed system on Cohn-Kanade (CK+), JAFFE, MMI and MUG benchmark databases to conduct and validate our experiment and established its performance superiority over other existing competitors.