Predicting future accident risks of older drivers by speech data from a voice-based dialogue system: A preliminary result

Abstract

As the world’s elderly population increases, driving accidents involving older adults has become an increasingly serious social problem. Previous studies have suggested cognitive impairments as one of the risk factors for future accidents. However, it remains unclear whether and how such future accident risks related to cognitive impairments can be predicted by using health monitoring technologies. In this study, we collected speech data from simulated conversations between 38 healthy older adults and a voice-based dialogue system. We followed up with the participants 1.5 years later and found that 17 of them had experienced near-accidents within the past year. We then built a binary classification model using the originally obtained speech data and found through leave-one-out cross-validation that it could predict whether a person would have a near-accident experience with 78.9% accuracy. Our preliminary results suggest that speech data from voice-based interaction systems might help older drivers recognize future accident risks.