Facial Micro-Expression Recognition Based on Deep Local-Holistic Network

A micro-expression is a subtle, local and brief facial movement. It can reveal the genuine emotions that person tries to conceal considered an important clue for lie detection. The research has attracted much attention due its promising applications in various fields. However, short duration low intensity of movements, recognition faces great challenges, accuracy still demands improvement. To improve efficiency feature extraction, inspired by psychological study attentional resource allocation cognition, we propose deep local-holistic network method recognition. Our proposed algorithm consists two sub-networks. first Hierarchical Convolutional Recurrent Neural Network (HCRNN), which extracts abundant spatio-temporal features. second Robust principal-component-analysis-based recurrent neural (RPRNN), global sparse features with micro-expression-specific representations. extracted effective are employed through fusion We evaluate on combined databases consisting four most commonly used databases, i.e., CASME, CASME II, CAS(ME)2, SAMM. experimental results show our achieves reasonably good performance.