细粒度分心驾驶行为识别方法OA

Distracted driving is one of the leading causes of traffic accidents,making the recognition of such behavior critical for improving road safety.However,existing graph convolution-based behavior recognition methods face significant limitations in capturing subtle and complex action features in driving scenarios.To address this,this paper proposes a fine-grained distracted driving behavior recognition network(FG-DDGCN)based on the Hierarchically Decomposed Graph Convolutional Network(HD-GCN).The model incorporates three key modules:Short-Term Explicit Motion Modeling(ST-EM),Multi-Scale Channel-Variable Spatial-Temporal Attention Mechanism(MSCVSTA),and Robust Decouple Loss(RDL),enabling precise recognition of subtle driver actions and significantly improving accuracy.The proposed model is evaluated on the Drive&Act data-set,and ablation experiments are conducted to validate the effectiveness of each module.On the fine-grained activity recognition task,the model achieves the best performance on both the validation and test sets(73.27%and 64.90%,respectively).In the atomic action unit task,the proposed model outperforms existing methods across the action,object,and location dimensions.Ex-perimental results demonstrate that FG-DDGCN exhibits promising capabilities in capturing fine-grained dynamic behavior fea-tures,achieving competitive performance compared to existing methods.This provides a feasible solution for distracted driving behavior recognition and offers valuable insights for future research in related fields.