An Efficient Illumination‐Invariant Dynamic Facial Expression Recognition for Driving Scenarios

• DOI: 10.1049/itr2.70009
• Publication Date: 2025-03-12T02:46:55Z
• Authors: Ercheng Pei; Man Guo; Abel Díaz Berenguer; Lang He; HaiFeng Chen
• Keywords: Facial expression recognition

ABSTRACT Facial expression recognition (FER) is significant in many application scenarios, such as driving scenarios with very different lighting conditions between day and night. Existing methods primarily focus on eliminating the negative effects of pose identity information FER, but overlook challenges posed by variations. So, this work proposes an efficient illumination‐invariant dynamic FER method. To augment robustness to illumination variance, contrast normalisation introduced form a low‐level features learningmodule. In addition, extract salient features, two‐stage temporal attention mechanism high‐level learning module deciphering facial patterns. Furthermore, additive angular margin loss incorporated into training proposed model increase distances samples categories while reducing belonging same category. We conducted comprehensive experiments using Oulu‐CASIA DFEW datasets. On VIS NIR subsets normal illumination, method achieved accuracies 92.08% 91.46%, which are 1.01 7.06 percentage points higher than SOTA results from DCBLSTM CELDL method, respectively. Based subset dark 91.25%, 4.54 result CDLLNet dataset, UAR 60.67% WAR 71.48% 12M parameters, approaching VideoMAE 86M parameters. The outcomes our validate effectiveness affirming its ability addressing diverse scenarios.