LCBP-STGCN: A local cube binary pattern spatial temporal graph convolutional network for micro-expression recognition.

This article focuses on the development of LCBP-STGCN, a novel method for automated micro-expression recognition that integrates Local Cube Binary Pattern (LCBP) features with a Spatial-Temporal Graph Convolutional Network (STGCN). The proposed approach constructs a spatiotemporal graph using Regions of Interest (ROI) as nodes and LCBP features as node information, employing Spatial Graph Convolutional Network (SGCN) and Temporal Convolutional Network (TCN) layers to extract high-level spatiotemporal features. Evaluated on four spontaneous micro-expression datasets—SMIC, CASME I, CASME II, and SAMM—LCBP-STGCN demonstrated superior recognition accuracy compared to traditional handcraft and several deep learning methods, achieving up to 81.26% accuracy on CASME I. The study highlights the method’s ability to capture subtle facial muscle movements by effectively combining spatial and temporal information, with future work aimed at cross-dataset evaluation and enhancing recognition based on Action Units (AUs).