A Body Area Network for Ubiquitous Driver Stress Monitoring based on ECG Signal

Abstract

During recent years, a body area network is becoming an important tool to improve the healthcare by monitoring patient's health state at distance, at home, in work, or when traveling. This body sensor-based network is easy to use, and available with low cost into two types; the first one can be swallowed or implanted under the skin where the second type is wearable, these became available within the Internet of Things (IoT). Physiological sensor-based systems have been recently designed to detect the emotional stress. By this way, we propose in this paper, a new monitoring system for driver stress detection based on an enhanced random forest classification approach. This proposal analyses and monitors driver electrocardiogram (ECG) signal when driving in order to discover its stress state belonging to one of the following three levels namely, low, medium or high. This proposal could help to detect and diagnostic stress level and alert the driver, its family and the other road users to avoid accidents caused by high stress state. The proposed system suggests the integration of a simulated annealing algorithm to enhance the random forest classification method in order to reach the highest classification accuracy. According to various drivers' ECG acquired from MIT-BIH physioNet dataset, the experimental study showed that the proposed random forest algorithm outperforms support vector machine (SVM) classification method to detect driver stress levels in terms of recognition accuracy.