An automatic approach to continuous stress assessment during driving based on fuzzy c-means clustering

This paper presents a novel approach for driving stress assessment by fuzzy clustering. In previous researches, stress during real-world driving tasks has been detected in discrete levels, but in this study, we demonstrated that considering fixed-levels for stress in long periods is not authentic. Without employing discrete levels of stress, data remains unlabeled. So a clustering method has been proposed to compensate for the lack of the feasibility of classification. Due to uncertainties, the clusters can be defined in terms of fuzzy sets. Furthermore, using fuzzy clustering methods, data overlap is considered. In the proposed algorithm, utilizing membership values generated by fuzzy c-means, and weights assigned by fuzzy inference system (FIS), we present automatic continuous criteria for stress in the short time intervals. The continuous scale is defined between 0and100, where higher values represent higher stress levels. Our findings not only confirm rough results of previous studies, but also indicate improvements in precision and accuracy of stress assessment.