Creating a Sri Lankan Micro-Emotion Dataset for a Robust Micro-Expression Recognition System

• Main Authors: Jayakodi, J. A. L. P., Jayamali, G. G. S. D., Hirshan, R., Aashiq, M. N. M., Kumara, W. G. C. W.
• Format: Conference Proceeding
• Language: English
• Subjects: action units; Computational modeling; Deep learning; emotion recognition; emotion stimulation; micro-emotion dataset; micro-emotion detection; Psychology; Solid modeling; Three-dimensional displays; Training
• ISSN: 2613-8662
• DOI: 10.1109/SCSE56529.2022.9905127

In interpersonal communication, the human face provides important signals of a person's emotional states and intentions. Furthermore, micro-emotions play a major role in understanding hidden intentions. In psychological aspects, detecting micro-emotions play a major role. In addition, lie detection, criminal identification, and security systems are other applications, where detecting micro-emotion accurately is essential. Revealing a micro-expression is quite difficult for humans because people tend to conceal their subtle emotions. As a result, training a human is expensive and time-consuming. Therefore, it is important to develop robust computer vision and machine learning methods to detect micro-emotions. Convolutional Neural Network (CNN) is the most used deep learning-based method in recent years. This research focuses on developing a 3D-CNN model to detect and classify Micro-emotions and creating a local Micro-emotion database. From the related research work we have considered this is the first attempt made at creating a Sri Lankan micro-emotion dataset. Having a local micro-emotion dataset is essential in formulating more accurate real-time applications focused on deep learning methods. Therefore, in this research, our main objective is to create a Sri Lankan micro-emotion database for future micro-emotion recognition and detection research.