Parallel Scale Invariant Feature Transform Based Approach for Facial Expression Recognition

Facial expression recognition has many important practical applications like Human-computer interaction, computer graphics, psychology etc. This paper proposes a parallel approach called Parallel SIFT (P-SIFT) for calculating Scale Invariant Feature Transform (SIFT) features which is then used for facial expression recognition. P-SIFT is also invariant to scale, rotation, blur, translation and partially invariant to illumination changes, 3D projection. In the proposed approach P-SIFT is used for key points detection, the results of which are used for facial expression recognition. One of the standard data set is used for testing the proposed approach. A database of Indian undergraduate student's images is also created with various expressions, which is also used for testing. The proposed P-SIFT approach shows increase in the computational speed. Results are encouraging and it takes almost 50% less time as compared to conventional SIFT. The experimental results show good prediction of emotions. The proposed approach successfully detects the emotions with variation in the levels of expressions.