Enhancing Driver Maneuver Intention Recognition: A Framework Integrating Driver Facial Motion and Driving Scene Understanding

Accurately identifying driver intentions is the cornerstone of human-centered autonomous driving assistance systems(ADAS). This paper presents a driver intention recognition approach that integrates driver facial information and driving scene information. The framework employs the driver facial motion (DFM) module and the driving scene understanding (DSU) module to to leverage internal and external features during the driver intention recognition process. The DFM is responsible for estimating driver facial key joints to generate spatiotemporal face graph and extracting driver feature. The DSU module provides lane information and driving scene feature embedding. The DFM and DSU modules work in parallel, and the outputs of both modules are fused and fed into a classifier to recognize driver intention. The proposed framework is evaluated on a public natural driving dataset and compared with state-of-the-art methods. Experimental results demonstrate superior performance of the proposed framework in recognizing driver intentions.