VGG 16 Based Human Emotion Classification Using Thermal Images Through Transfer Learning

The monitoring of human physical as well as mental health have gained importance in today's stressful world. Human Emotions play an important role in measuring and evaluating mental state of human beings. The paper is here to present the novel way to analyze human emotions using thermal imaging. The thermal images from 36 subjects are acquired. The six primary emotions plus neutral state is classified here. The thermal images are acquired using thermal imager FLUKE TiS20. The subjects are in the age group 21 years to 57 years, who are working/ studying in the Malnad College of Engineering, Hassan, Hassan district in Karnataka, India. A pretrained Convolutional Neural Network (CNN), VGG 16 has been used to extract features and to classify primary emotional states. The proposed methodology yielded the average recognition rate of 0.28.