Drivesafe: Pose-Based Driver Monitoring In Software Defined Vehicle Using STGCN

Driver state monitoring is crucial in the realm of software-defined vehicles. This monitoring ensures a continuous evaluation of driver attentiveness to prevent potential accidents resulting from distractions or fatigue. This technology plays a significant role in the ever-evolving landscape of autonomous and semi-autonomous vehicles, fostering a safer and more reliable driving experience. In our proposed methodology, the first process is to extract the driver’s skeletal features from RGB video captured by a camera. We extract the skeleton feature by utilizing the OpenPose method. Subsequently, we construct the graph by considering the key joints of the body as a node, followed by applying the Spatial-Temporal Graph Convolutional Network (STGCN) for action recognition. Our approach’s effectiveness is substantiated by using the proposed method on the TiCAM dataset. This application shows that our method has the potential to improve the monitoring of driver states and contribute to the progress of softwaredefined vehicles.