Micro Expression Detection Using Deep Learning Techniques

Identifying emotions from a person's face which lasts for few seconds has been done by several researchers in the past. A micro expression is a brief facial expression that occurs when the amygdala responds rapidly to stimuli, reflecting the individual's attempt to hide specific emotions they are experiencing. Micro expressions could be used to analyse the deeper emotions of a person as it is expressed naturally. In this work, multiple models of neural network and pre-trained with default parameters is proposed. Two micro expression datasets were used in this work (Cohn-Kanade and FER 2013.) The classification algorithms used here are neural network models which are Convolution Neural Network (CNN), RESNET 50, VGG-19, VGG-16, INCEPTION V3 and Efficient net B0. Feature Extraction and training is done using these models and performance evaluation is done. For CK+, an accuracy of 99% is obtained and for FER 2013, 95% is achieved.