Estimation of Driver's Attention Level Based on Correlation Analysis of Movements

The take-over process from automatic to manual control plays a critical role in safety insurance of the Level 3 automated driving, and driver's state prior to the handover transition process has a great influence on the take-over quality. Estimation of driver's engaging level in various non-driving activities (NDAs) is therefore of great importance to the design of the take-over control strategy. In this paper, focusing on a specific NDA of playing tablet, a new approach to estimate the driver's attention level by calculating the time-varying correlation coefficient between driver's head movement and tablet's movement, as well as its fluctuation. The experiment with an operator paly a tablet is conducted. The experimental results show that the correlation coefficient has a significant positive correlation to the attention level and it can be adopted as a feature to represent driver's state in NDAs. The results demonstrate the feasibility and effectiveness of the proposed method and also presents a good potential to be used in real vehicle application of designing take-over control algorithms.