I. Introduction
Facial behavior analysis is one of the fastest growing research areas in affective computing and computer vision research. We can learn about people’s emotions through facial behavior. There are two commonly used ways to describe facial behavior: 1) facial expression and 2) facial action unit (AU). Facial expression is an intuitive description of facial behavior, most commonly identified as one or more of six expressions (i.e., anger, disgust, fear, happiness, sadness, and surprise) [1]. These labels, however, are not complex enough to describe the full range of emotions. There are many other expressions, including pain, awe, embarrassment, hatred, etc. The number and definition of expressions are not universally agreed upon by researchers.