Development of novel deep learning-based facial emotion recognition for atypical individuals

Abstract

Atypical individual facial expressions are hard to understand because their expressions are different from normal people's. They have a lower expressive quality in their facial expressions. Facial emotion recognition system (FER) can recognize an atypical person's emotions automatically. Researchers struggle with the dataset's small size and difficulty getting facial data from the atypical person. We have investigated several ways to further improve FER. Also, we have recognized that atypical people have difficulty understanding the emotions of normal people. We want to develop an app or software to assist atypical individuals in recognizing normal people's emotions, and it is helpful for atypical individuals to learn other people's emotions. We looked at previous studies of neutral expression and proposed a novel convolutional neural network-based FER for recognizing neutral expression and expressive emotion.