Facial Expression Recognition Based on 2D Gabor Transforms and SVM
In the facial expression recognition, a dimension disaster will arise when taking the coefficient of Gabor transforms as the expression eigenvectors. To avoid this issue we draw grids on facial region, making the mean coefficient value of Gabor transforms of each gird as the eigenvectors. Furthermore we classify the expression by constructing the multi-class C-SVC, improved the accuracy and speed of the algorithm by dropping the redundant features using sequential backward selection. The experimental result proves the superiority of the algorithm we proposed to other algorithms.
C. H. Liu et al., "Facial Expression Recognition Based on 2D Gabor Transforms and SVM", Applied Mechanics and Materials, Vols. 58-60, pp. 238-242, 2011