OpenDriver: An open-road driver state detection benchmark

Wearable physiological measurements offer a convenient and feasible method for real-time driver states monitoring. However, there are currently few driver physiological datasets in open-road scenarios, and the existing datasets suffer from issues such as poor signal quality, small sample sizes, and short data collection periods. In this paper, a large-scale multi-modal driving benchmark namely OpenDriver is elaborately constructed for driver state detection. Firstly, the OpenDriver encompasses 3278 driving trips, with a signal duration of approximately 4600 h. Two modalities of driving signals are collected: electrocardiogram (ECG) signals and six-axis motion data of the steering wheel from a motion measurement unit (IMU), which are recorded from 81 bus drivers and their vehicles. Secondly, three challenging tasks are carefully designed, and they are ECG signal quality assessment, individual biometric identification based on ECG signals, and physiological signal analysis in complex driving environments, respectively. Moreover, the corresponding baseline models and evaluation metrics are proposed to demonstrate the rationality and completeness of the dataset and tasks. First, in the quality assessment task, a noisy augmentation strategy is introduced to achieve realistic noise simulation, and then a larger-scale ECG signal dataset is generated. Second, an end-to-end contrastive learning framework is employed to effectively identify individual biometric. Finally, a comprehensive analysis of drivers’ Heart Rate Variability (HRV) features under different driving conditions gives multiple heuristic analytical conclusions. The OpenDriver benchmark and dataset will be publicly available at https://github.com/bdne/OpenDriver.