Towards smart helmet for motorcyclists: automatic stress level detection using wearable accelerometer sensor system

In the modern world, stress is a common occurrence that alters how people behave and how they physically react to certain situations. However, Consistent or prolonged stress can put our well-being at risk and disrupt our everyday tasks. Assessing stress and other psychological factors in humans is difficult because it involves self-reporting, which can be subjective, and there can be differences both among and within individuals. The advancement of Internet of Things (IoT) sensors has enabled the tracking of a wide range of human behavior patterns, including stress-related behavior that can be quantified objectively. This study makes use of a wearable accelerometer sensor that was used to collect information on a real-world scenario concerning a motorcyclist’s mental health. In this study, a wearable single triaxial accelerometer sensor was placed inside the motorcycle rider’s helmet to collect data on his physiological and behavioral characteristics. Using various machine learning methods such as Decision Trees (DT), Discriminant Analysis (DA), Support Vector Machines (SVM), Nearest Neighbor Classifiers (k-NN), and Ensemble Classifiers, the accuracy of motorcyclist stress was assessed and compared. By employing classification methods, the Quadratic Support Vector Machine (QSVM) exhibited the highest average accuracy of 83.3% and an average F-score of 91.3% overall. In contrast, the Quadratic Discriminant (QD) analysis demonstrated the lowest average accuracy of 62.5% and an average F-score of 70.7%. A questionnaire that measured privacy concerns on a scale of 1- 5 was used. The findings indicate that our approach has the potential to deliver precise identification of stress levels among motorcyclists, hence demonstrating its usefulness for future adoption.