Novel Method of Stress Detection Using Physiological Measurements of Automobile Drivers

Abstract

Driving a car is a complex cognitive process in which even a small lack of attention can have disastrous consequences. In order to minimize human error while driving, we can monitor stress and fatigue by measuring physiological parameters like ElectroCardioGram (ECG), ElectroMyoGram (EMG), Skin Conductance (SC) also called as Galvanic Skin Response (GSR) and Respiration Rate (RR) continuously over a period of time. Autonomic Nervous System (ANS) primarily depends on emotional responses of the human body to the dynamic surrounding. Further it also controls the smooth muscles, heart muscle and secretion of the glands in human body. As a result of this fact, bio-signal recordings reflecting the operating condition of the physiological systems including the circulatory, respiratory, muscular and endocrine systems can provide useful information representing the dynamics of the internal states in human body. Hence, the dynamic mental stress level of an automobile driver can be derived from those recordings. In this research we accessed raw physiological signals available at PHYSIONET website and then extracted useful statistical features. Correlation analysis on the selected features showed that Mean HR and Mean Hand GSR are the two statistical features that have a very strong correlation with changing traffic conditions. We presented a method based on a correlation analysis and developed a mathematical function for the estimation of automobile driver stress level. The proposed methodology monitors driver’s stress level using features extracted from selected physiological parameters. The results obtained indicate a strong correlation between the stress level of driver and the mathematical function formed. We used threshold approach to perform classification of affective states as “Low Stress”, “Moderate Stress” and “High Stress” based on different traffic conditions. The results indicate classification accuracy of more than 80% in most of the driver data sets. Thus, the stress function acts as a direct indicator of stress level of the automobile diver whose physiological parameters are monitored continuously in real time.