Recognizing facial expression based on discriminative multi-order Two Dimensions Fractional Fourier Transform

Two Dimensions Fractional Fourier Transform (2D-FrFT) is a powerful tool for feature extraction in facial expression recognition. However, a single order 2D-FrFT feature is not effective enough. In this paper, it proposes a facial expression recognition system based on multi-order with 2D-FrFT features. Moreover, a multi-feature fusion algorithm which is named Discriminative Multi-set Canonical Correlation Analysis (DMCCA) is deduced. Fusion of multi-order 2D-FrFT features by DMCCA can significantly improve the recognition performance, due to increased amount of discriminative information. More important, an objective function is defined to find the optimal 2D-FrFT orders group fused by DMCCA. Experiment results demonstrate the effectiveness of the method.