Characterization and Identification of Driver Distraction During Naturalistic Driving: An Analysis of ECG Dynamics

One of the most contributing factors to the accidents on the roadways is distracted driving. While in-vehicle, driver may get distracted by variety of ways such as talking on the cellphone, conversing with the accompanying passengers, texting while driving, etc. In order to reduce potential chances of road-accidents, it is highly essential to characterize and identify distracted situations in real-time. In this paper, we investigate Electrocardiogram (ECG) signals as the physiological measure to characterize driver distraction. We aim to provide an empirical guideline for accurate and in real-time analysis irrespective of the body physic. ECG-based driver distraction identification has significant advantages in practice such as being easy to capture, minimally intrusive, and reliable in biometric patterns. ECG dynamics encompass multiple descriptors that we examine in this investigation for efficient characterization of driver state toward real-time identification of distracted driving. In this effort, six drivers were actively participated in our naturalistic driving experiments, where the distraction is introduced by the cellphone conversation and the conversation with the passenger. Our study mainly focuses on the efficient characterization of distraction by localizing R-R interval based on temporal features as well as spectral features. In addition to this, we further investigated the real-time predictive ability of the extracted features through state of the art predictive algorithms. Our experimental results demonstrated ∼83% average predictive accuracy of driver distraction identification in near real-time.