Low dimensional feature vector based on the combination of fractional fourier transformand graph signal processing for facial expression recognition
Hemant Kumar Meena, Kamalesh Kumar Sharma, Shiv Dutt Joshi · 2017 IEEE International Conference on Power, Control, Signals and Instrumentation Engineering (ICPCSI) · 2017
Fractional Fourier Transform (FRFT) as a time-frequency analysis tool has been used for the facial expression recognition. However, the major problem is that the length of the feature vector is quite large. We have proposed the use of the graph signal processing (GSP) to reduce the dimension of the FRFT feature vector. Compared to the FRFT, the accuracy of the facial expression recognition rate increases by the combined FRFT-GSP method. Experimental results on CK+ database demonstrates the effectiveness of the proposed work.