Facial Emotion Recognition in Curvelet Domain

The work being presented here describes a novel approach for human emotion recognition based on curvelet transform. Our approach of the emotion recognition is motivated by the fact that emotions expressed more obviously by the facial curves, hence the technique proficient in capturing the edge singularity as well as the curve singularities shall yield the better results i.e. curvelets. Curvelet coefficients were obtained by applying discrete curvelet transform on the facial expression image set JAFFE. The features were calculated by applying common statistics and classification was performed using support vector machine (SVM). The results obtained are promising, leading towards the inference that proposed method is more effective for the emotion recognition problem than the other comparable methods. Our approach was also applied on another In-House dataset and performance is being shown in result.