Micro-Expression Recognition by 3D-CNN in Combination with GRU

Chun-Ting Fang, Tsung-Jung Liu, Kuan-Hsien Liu

Abstract: Micro expressions are fleeting emotional expressions that manifest within a brief time frame, often characterized by subtle facial muscle movements. Despite their brevity and imperceptibility to the naked eyes, these micro expressions encapsulate valuable emotional cues. The objective of micro expression recognition is to leverage AI technology to effectively detect and analyze these nuanced facial changes, thereby enhancing our ability to gain deeper insights into an individual’s underlying emotional state. In this paper, we introduce an approach featuring a three-layer 3D-CNN architecture combined with GRU. Experimental results demonstrate that our method yields favorable outcomes despite its relatively straightforward architecture.