Recognizing smile emotion based on Fractional Fourier Transform

Fractional Fourier Transform (FRFT) is a time-frequency analysis tool which contains the time-frequency information of the signal at the same time. In this paper, we use FRFT as feature extraction method for recognizing smile emotion. Based on the FRFT, the original signal is transformed into complex-values containing amplitude and phase information. We adopt to use amplitude, phase and complex information as feature extraction for smile recognition, respectively. Moreover, the AdaBoost algorithm was used for classification. In this paper, experiment results simulating on datasets called RML expression database and JAFFE database demonstrated that FRFT features could achieve as a high recognition rate as the Gabor features did, which showed that this approach was efficient in some degree.