Facial Micro-expressions Analysis: Its Databases, Feature Extraction, and Classification Methods

Abstract

Facial micro-expressions (ME) are a subtle expression that appears within a shorter duration compared to facial macro-expressions. This expression is a type of universal non-verbal communication that appears on someone’s face unconsciously when he or she is attempting to hide their true emotions. It reveals a person’s true emotional state, which is vital for both mental health diagnosis and security purposes. Because micro-expressions are typically low in intensity and short in duration, humans and/or machine learning models can be specifically trained to recognize some expressions, but the obtained performance is usually consistently poor. The research on facial micro-expression recognition is relatively new in the field of computer vision. This paper presents an overview of facial micro-expressions and their taxonomy. Each posed and spontaneous database, as well as feature representation and classification techniques, are examined and expanded upon. In addition, we conclude the overview by pointing out a few flaws in micro-expression recognition methods and making some recommendations for future research directions that could aid researchers in making progress in this field.