Stress Detection Through Electrodermal Activity (EDA) and Electrocardiogram (ECG) Analysis in Car Drivers

R. Rinaldo,Antonio Affanni,R. Bernardini,A. Piras,P. Zontone

DOI: https://doi.org/10.23919/EUSIPCO.2019.8902631

2019-09-01

Abstract:The stress in a driver, happening during unforeseen events or taxing situations, is linked to a subject’s sympathetic system response. We present a system which detects the stress presence in car drivers through the analysis of an endosomatic Electrodermal Activity (EDA) signal, namely, Skin Potential Response (SPR), coupled with the analysis of the Electrocardiogram (ECG) signal. To log these signals we utilize a device which records the SPR from each hand of the driver, and the ECG from the chest. In the case of the SPR signal, since the hands movement injects motion artifacts, we also utilize an algorithm that dynamically selects the smoother signal coming from the two hands, and is thus able to output a clean SPR signal. Statistical features are then derived from the ECG and SPR signals, allowing their classification using a Supervised Machine Learning Algorithm. Various subjects were tested in an environment set in a company which develops professional driving simulators, both in hardware and software, and consisted in a motorized platform, a cockpit and a 180° projection screen. The test encompassed driving through a highway, with some unforeseen events happening at some positions. In the end we get a Balanced Accuracy in stress detection of 77.59 % for the considered events.

Medicine,Engineering,Computer Science