Comparative Study on Facial Expression Recognition using Gabor and Dual-Tree Complex Wavelet Transforms

Authors : Alaa Eleyan

Pages : 1-13

Doi:10.24107/ijeas.283277

View : 14 | Download : 11

Publication Date : 2017-04-07

Article Type : Research Paper

Abstract :Moving from manually interaction with machines to automated systems, stressed on the importance of facial expression recognition for human computer interaction insert ignore into journalissuearticles values(HCI);. In this article, an investigation and comparative study about the use of complex wavelet transforms for Facial Expression Recognition insert ignore into journalissuearticles values(FER); problem was conducted. Two complex wavelets were used as feature extractors; Gabor wavelets transform insert ignore into journalissuearticles values(GWT); and dual-tree complex wavelets transform insert ignore into journalissuearticles values(DT-CWT);. Extracted feature vectors were fed to principal component analysis insert ignore into journalissuearticles values(PCA); or local binary patterns insert ignore into journalissuearticles values(LBP);. Extensive experiments were carried out using three different databases, namely; JAFFE, CK and MUFE databases. For evaluation of the performance of the system, k-nearest neighbor insert ignore into journalissuearticles values(kNN);, neural networks insert ignore into journalissuearticles values(NN); and support vector machines insert ignore into journalissuearticles values(SVM); classifiers were implemented. The obtained results show that the complex wavelet transform together with sophisticated classifiers can serve as a powerful tool for facial expression recognition problem.
Keywords : facial expression recognition, complex wavelet transform, local binary pattern, principle component analysis, neural networks, support vector machines