Extracting tendency and stability from time series and random forest for classifying a car driver´s cognitive load

This paper presents a novel temporal feature extraction method and random forest (RF) for classifying a car driver´s cognitive load. Temporal value, tendency, and stability are important features for classifying a car driver´s state. We present the classification problem of the car driver´s state. We need a function in the in-vehicle information service that judges the user´s cognitive load. We define the driver´s cognitive load based on the driving situation. The experiment confirmed that classification accuracy improved using the tendency and stability of a time series. Moreover, the results confirmed that the tendency and stability of steering angle, accelerator rate, and car speed contribute to the classification of cognitive load.