dcyphr | Electrodermal Activity Sensor for Classification of Calm/Distress Condition


This article introduces a novel unobtrusive wearable device that measures electrodermal activity (EDA) for health related systems. EDA measures the changes in conductivity in the skin because of increases or decreases in sweat gland activity. This device is placed on the wrist and continuously monitors if the user is in a calm or distress condition. The results indicate 89% accuracy when using only data from the EDA signal. 


The detection of stress can prevent many health issues related to distress. This warrants the need for a device to monitor and interpret a person's arousal state which would be beneficial for maintaining a person's emotional well being. EDA is directly related to the sympathetic nervous system, which works to cause arousal. Similarly, EDA sensors are cheap, small, and light weight which make it a prime candidate for implementation into a wearable. The best place for EDA sensors is the palms of hands or soles of the feet.


Participants sit in front of a monitor with the wearable placed onto the wrist on the non-dominant hand. 10 neutral or emotionally charged images were shown in series to participants for 6 seconds with 1 second black screens in between. Before switching from emotional to non-emotional images, a 10 second black screen was shown to bring the participants to a neutral emotional state.


Higher values of skin conductivity measured from EDA are observed during distress. 


Other systems that measure stress use a variety of other physiological and camera sensors to reach an accuracy of 68%. These systems achieve a worse outcome with more sensors. The researchers method uses only the measurement of EDA and is notable due to its small hardware size, signal processing capability, and software accuracy. However, this study is limited in several ways. There were only 45 participants all of which were young, which reduces generalizability. Similarly, showing images within the same emotional category in series may affect the outcomes and could suppose an experimental bias. Finally, no other stimulus besides images was tested.