One of the most important characteristics of human emotion is micro-expression. This micro-expression can be grouped based on several classes such as Contempt, Fear, Happy, Tense, Disgust, Repression, Surprise, and Sad. Even though micro-expression being one of the major research interest areas, there is a lot of scope for improvement in terms of micro-expression to emotion mapping. The primary objective of this paper is to infer the emotions of humans based on their micro-expressions using deep learning techniques. In this paper, videos from standard micro-expression benchmarked databases like CASME I,SMIC-HS, SMIC-NIR, SMIC-VIS, and CASME II were collected and parsed into images and then trained and tested using an Attention embedded Residual Network model. This model brings a novel method of embedding Attention on Residual Convolution Neural Networks, which results in carrying the most significant dominating features till the very end. The proposed work on these microexpression datasets yields better results with a state-of-the-art accuracy of more than 95% in each dataset. For verification purposes, the videos collected real-time manually from different gender people were tested and an accuracy score of 87.37% and an F1_score of 86.92% was achieved.
Volume 12 | 07-Special Issue
Pages: 1261-1272
DOI: 10.5373/JARDCS/V12SP7/20202226