Driver's stress detection using Skin Potential Response signals. (July 2018)

• Record Type: Journal Article
• Authors: Affanni, Antonio; Bernardini, Riccardo; Piras, Alessandro; Rinaldo, Roberto; Zontone, Pamela
• Is Part Of: Measurement. Volume 122(2018)
• Page Start: 264
• Page End: 274
• Publication Date: 2018-07
• Subjects: Stress detection -- Electrodermal activity -- Skin potential response -- Motion artifact -- Adaptive filters; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2018.03.040
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: We analyze the problem of automatic detection of car drivers' stress levels. Stress levels are detected from the driver's hand Skin Potential Response signals. Artifacts due to the hand movements are removed using adaptive filters. Stress events are detected as peaks of a Smoothed Nonlinear Energy Operator output. We obtain Recall values with induced stress events as high as 95%. Abstract: The problem of automatic detection of car drivers' stress levels has become increasingly important, due to its impact on people security, and more generally on people health and well-being. Among the various techniques proposed for stress detection, Electrodermal Activity (EDA) monitoring is particularly interesting to gain information about the inner stress affecting a person, due to its correlation with the sympathetic nervous system response. In the application to driver's stress detection, EDA parameters are strongly affected by Motion Artifact caused by physical movements of the subject under test. In this paper, we propose a scheme based on EDA Skin Potential Response (SPR) measurements, together with records of the steering wheel angle, which is used in an adaptive filter setup to remove motion artifacts. We also show that, by appropriately processing EDA/SPR signals only, it is possible to efficiently locate stress events during driving. We then propose an experimental setup which allows defining a ground-truth for stress events recognition, and which confirms the validity of the proposed approach.