Compound Micro-Expression Recognition System

Micro-expressions are facial movements with subtle amplitude and short duration. Similar to normal expressions (i.e., macro-expressions), micro-expressions correspond to six basic emotional categories: happiness, sadness, surprise, fear, anger, and disgust. However, a large number of psychological studies show that people will produce and use many more complex expressions in their daily life, called "compound expressions". Compound expressions are composed of the six basic expressions but reflect more complex mental states and more abundant human facial emotions. In this paper, we introduce the concept of compound expression into the detection and recognition of micro expression for the first time and generated the Compound Micro-Expression Database (CMED). In addition, this paper will extract the optical flow feature map of the onset frame and the apex frame in the compound micro-expression sequences and input it into the pre-designed shallow convolutional neural network. The proposed method synthesized five existing databases of spontaneous micro-expressions (CASME I, CASME II, CAS(ME)^2, SMIC, SAMM) to generate the CMED and test the validity of our network. The experimental results show that the deep network framework designed in this paper can well describe and recognize the emotional information of micro-expression and compound micro-expression.