Development of an emotion recognition system based on face images
Loading...
Date
2017-06
Authors
Tan, Yi Chen
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Automatic facial expression recognition has vast applications such as sociable robots, intelligent tutoring system, smart home automation system and other human-machineinteraction applications. Thus, the research in facial expression recognition has been growing in interest. However, most approaches do not compare the impact of data augmentation methods and the overall error rate is still high. In this project, a facial expression recognition system based on Convolution Neural Network is proposed. The system extracts all relevant information (i.e., features) from two-dimensional digital facial image and classifies the image into one of the universal facial expression (i.e. happiness, surprise, sadness, disgust, fear, anger and neutral). To increase the accuracy of the proposed system, a number of pre-processing are carried out. The training data are augmented by using methods such as adding salt-and-pepper noises, Gaussian noise, brightness variations and flips. The system is evaluated by using widely used JAFFE and CK+ databases and methods such as k-fold cross-validation and cross-database validation. The proposed method achieved good result, which is 84.06% using CK+ database and 77.59% using combined data from CK+ and JAFFE databases. The system achieved the highest accuracy using data augmented with flips, which is increased from 74.40% to 77.17%. Therefore, data augmentation is proven that it can increase the accuracy.