Affiliations: [a] Informatics Department, University of Oviedo, Campus de Gijon, Calle Fojanes, s/n, 33204 Gijon, Spain. E-mail: [email protected] | [b] Telematics Department, Universidad Carlos III de Madrid, Av. de la Universidad, 30, 28911 Leganés, Spain
Abstract: The number of motorcycles on the road has increased in almost all European countries according to Eurostat. Although the total number of motorcycles is lower than the number of cars, the accident rate is much higher. A large number of these accidents are due to human errors. Stress is one of the main reasons behind human errors while driving. In this paper, we present a novel mechanism to predict upcoming values for stress levels based on current and past values for both the driving behavior and environmental factors. First, we analyze the relationship between stress levels and different variables that model the driving behavior (accelerations, decelerations, positive kinetic energy, standard deviation of speed, and road shape). Stress levels are obtained utilizing a Polar H7 heart rate strap. Vehicle telemetry is captured using a smartphone. Second, we study the accuracy of several machine learning algorithms (Support Vector Machine, Multilayer Perceptron, Naïve Bayes, J48, and Deep Belief Network) when used to estimate the stress based on our input data. Finally, an experiment was conducted in a real environment. We considered three different scenarios: home-workplace route, workplace-home route, and driving under heavy traffic. The results show that the proposal can estimate the upcoming stress with high accuracy. This algorithm could be used to develop driving assistants that recommend actions to prevent the stress.
