Localized discriminative scale invariant feature transform based facial expression recognition

This paper presents a discriminative scale invariant feature transform (D-SIFT) based feature representation for person-independent facial expression recognition. Keypoint descriptors of the SIFT features are used to construct distinctive facial feature vectors. Kullback Leibler divergence is used for the initial classification of the localized facial expressions and weighted majority voting based classifier is employed to fuse the decisions obtained from localized rectangular facial regions to generate the overall decision. Experiments on the Bosphorus and BU-3DFE databases illustrate that the D-SIFT is effective and efficient for facial expression recognition.

Graphical abstractFigure optionsDownload full-size imageDownload as PowerPoint slideHighlights► D-SIFT requires detection of localized descriptors to improve feature matching. ► We chose 4 × 4 uniform grids and top two discriminating descriptors for each sub-region. ► We utilize the descriptors based on Fisher criterion. ► WMV-based classifier is employed to fuse the decisions obtained from each sub-region.