Facial Emotion Recognition Based on Two-dimensional Empirical Mode Decomposition and Pca plus Lda

This paper proposes a new approach of using nonlinear technique, two-dimensional empirical mode decomposition (2DEMD) and PCA plus LDA for facial emotion recognition. The EMD is a non-parametric data-driven analysis tools which decomposes any nonlinear and non-stationary signals into a number of intrinsic mode functions (IMFs). In this work we used the 2DEMD which is the extension of one dimensional EMD to extract the features at multiple scales or spatial frequencies from facial images. These features called IMFs that obtained by a sifting process. To reduce dimensional features, PCA plus LDA was applied on IMF features. The obtained features were classified using knearest neighbor classifier. To evaluate the effectiveness of the proposed method, Cohn-Kanade database was employed. A series of experiment shows that the proposed method achieves recognition rate of 98.28% thus demonstrates a promising result for classifying the facial emotions.