Training computers to recognize facial expressions of humans: a machine learning approach

Abstract

Humans are precise in facial expressions and recognizing them. Facial expression is a routine task of humans. Training Computers to recognize such humanly task as closed to humans can be useful in numerous ways such as offering services as per the mode of consumer or user and will enhance the phenomenon of human computer interaction. Environment and surroundings are the sources from where humans naturally learn the models of facial expressions and they have logical models in brain, and with the comparison of models, they are able to recognize the expressions of others. This research aimed to design and develop a robust facial expression recognition system by combining various techniques available in Computer Vision and Machine Learning. Hence, it was crucial to understand the human psychological aspects of facial expression in a precise manner. For the training and demonstration purposes, human volunteers were utilized along with facial expression dataset such as CK+ from internet. Local Binary Patterns (LBP) and Convolutional Neural Network (CNN) were occupied for feature extraction and optimization respectively. In addition, different kinds of AI based tools and techniques were employed to mimic the human ability of recognizing facial expressions. The outcome presents a novel approach to facial expression recognition with modified LBP and CNN. The outcome of the research suggests that further studies and optimization would lead to commercially viable solutions such as products or services as per the expression and emotions of the consumers. The researchers will continue to optimize the outcome to supersede humans in recognizing facial expressions in future