Pen-type electrodermal activity sensing system for stress detection based on likelihood ratios



Journal Title

Journal ISSN

Volume Title



Psychological stress experienced during academic testing is known to be a significant performance factor for some students. While a student may be able to recognize and self-report stress experienced during an exam, unobtrusive tools to track stress in real time (and in association with specific test problems) are lacking. This effort pursued the design and initial assessment of an electrodermal activity (EDA) sensor - essentially a sweat sensor - mounted to a pen/pencil 'trainer:' a holder into which a pen/pencil is inserted that can help a person learn how to properly grip a writing instrument. This small assembly was held in the hand of a given subject during early human subject experiments and can be used for follow-on, mock test-taking scenarios. Data were acquired with this handheld device for 37 subjects (Kansas State University Internal Review Board Protocol #9864) while they each viewed approximately 30 minutes of emotion-evoking videos. Data collected by the EDA sensor were processed with an EDA signal processing app, which calculated and stored parameters associated with significant phasic EDA peaks. These peak data were then evaluated by a hypothesis driven stress-detection test that employed an approach using likelihood ratios for the ‘relaxed’ and ‘stressed’ groups. For these initial, motion-free testing scenarios, this pen-type EDA sensing system was able to discern which phasic responses were associated with ‘relaxed’ versus ‘stressed’ responses with 85% accuracy, where subject self-assessments of perceived stress levels were used to establish ground truth.



Electrodermal Activity, Stress detection, Likelihood ratios, Generalized extreme value distribution, Academic emotions, Classification

Graduation Month



Master of Science


Department of Electrical and Computer Engineering

Major Professor

Steven Warren