The research in the field of visual emotion understanding and analysis primarily encompasses micro-expression video recognition and localization, emotionally tinted image/video content description, and facial expression recognition and analysis.

Micro-expression video recognition and localization is a purely visual understanding challenge. Its main focus is on detecting and segmenting micro-expressions in videos, and further inferring the internal emotional changes of the people in the video based on these micro-movements.

Emotionally tinted visual content description is a branch of research that evolved from traditional visual content description. Its primary objective is to extract factual and emotional tendencies from visual content and use these emotional tendencies to describe the facts. For example, while traditional visual content description might simply state "a boy is telling a story to his classmates," an emotionally tinted visual description would elaborate it as "the boy is happily sharing a story with his classmates."

Facial expression recognition and analysis is an extension of micro-expression video recognition. It involves identifying changes in facial expressions from facial videos and making emotional inferences. Additionally, this approach can predict common physiological signals (such as blood pressure, heart rate, etc.) by focusing on the visual changes in human facial skin.