Improved Facial Expression Recognition Based on DWT Feature for Deep CNN

Electronics

Abdelmalik Taleb-Ahmed¹  Mohammed Beladgham²  RidhaIlyas Bendjillali²  Khaled Merit² 

[1] Laboratory of IEMN DOAE. UMR CNRS 852, University of Valenciennes, 59313 Valenciennes, France;Laboratory of TIT, Department of Electrical Engineering, Tahri Mohammed University, Bechar 08000, Algeria;

关键词: facial expression recognition (FER);    deep convolutional neural network (deep CNN);    discrete wavelet transform (DWT);    contrast limited adaptive histogram equalization (CLAHE);   

DOI  :  10.3390/electronics8030324

来源: DOAJ

【 摘 要 】

Facial expression recognition (FER) has become one of the most important fields of research in pattern recognition. In this paper, we propose a method for the identification of facial expressions of people through their emotions. Being robust against illumination changes, this method combines four steps: Viola–Jones face detection algorithm, facial image enhancement using contrast limited adaptive histogram equalization (CLAHE) algorithm, the discrete wavelet transform (DWT), and deep convolutional neural network (CNN). We have used Viola–Jones to locate the face and facial parts; the facial image is enhanced using CLAHE; then facial features extraction is done using DWT; and finally, the extracted features are used directly to train the CNN network, for the purpose of classifying the facial expressions. Our experimental work was performed on the CK+ database and JAFFE face database. The results obtained using this network were 96.46% and 98.43%, respectively.

【 授权许可】

Unknown