Psychological Characteristics Estimation from On-Road Driving Data.

Automobiles are essential to society, while they cause some problems. In particular, traffic accidents by older drivers are a serious problem. Driving assistance systems are promising solutions to this problem. However, it is not easy to develop acceptable and comfortable driving assistance systems for all drivers. To realize such systems, we need to consider the driver characteristics or personality of the drivers. In this paper, we predict the psychological characteristics of older drivers from on-road driving data and propose a classification model. We posit that road types are important information for estimation and that important driving behaviors appear not only in whole driving but also in partial driving. Under these hypotheses, our feature extraction method segments time-series data using road types and further segments data into various duration sequences. Experimental results show that some items can be predicted with high accuracy and validate the efficacy of the segmentation. We use a dataset that includes time-series driving data, the Driving Style Questionnaire scores, and the Workload Sensitivity Questionnaire scores of 24 older drivers. Also, this study gives a new perspective to the prediction of individual characteristics.