Toward Collecting Driving Data Using Wrist-Worn Sensors in Real Vehicles and its Application to Driver Activity Recognition

Abstract

To prevent traffic accidents caused by human factors, it is essential to identify the potential risk factors in daily driving. To address this challenge, a low-cost and easy-to-use method is required to continuously collect and measure the driver’s state information while protecting the driver’s privacy. To address this issue, we have developed a driver state estimation method using wrist-worn sensors. However, the verification of our approach has been limited to laboratory environments thus far. Therefore, in this study, we developed an in-vehicle measurement system using wrist-worn sensors and conducted experiments to collect driving behavior data in a real-world environment. We then estimated the driver’s behavior from the collected hand acceleration data. From the results, we identified challenges in applying our method to real-world conditions. Finally, based on these approaches, we developed a framework for collecting data in real-world environments and constructing datasets.