Facial Emotion Recognition Based on Empirical Mode Decomposition and Discrete Wavelet Transform Analysis

Abstract

This paper presents a new framework of using empirical mode decomposition (EMD) and discrete wavelet transform (DWT) with an application for facial emotion recognition. EMD is a multi-resolution technique used to decompose any complicated signal into a small set of intrinsic mode functions (IMFs) based on sifting process. In this framework, the EMD was applied on facial images to extract the informative features by decomposing the image into a set of IMFs and residue. The selected IMFs was then subjected to DWT in which it decomposes the instantaneous frequency of the IMFs into four sub band. The approximate coefficients (cA1) at first level decomposition are extracted and used as significant features to recognize the facial emotion. Since there are a large number of coefficients, hence the principal component analysis (PCA) is applied to the extracted features. The k-nearest neighbor classifier is adopted as a classifier to classify seven facial emotions (anger, disgust, fear, happiness, neutral, sadness and surprise). To evaluate the effectiveness of the proposed method, the JAFFE database has been employed. Based on the results obtained, the proposed method demonstrates the recognition rate of 80.28%, thus it is converging.