Improving Human Behavior Recognition Using Bilateral Electrodermal Activity Data Collected From Wearable Devices

Wearable devices, such as smartwatches and smartrings, are increasingly used to collect physiological signals, including electrodermal activity (EDA), which serves as a proxy for autonomic nervous system (ANS) activation. Prior research has shown that brain lateralization — where each hemisphere is associated with specific functions — affects EDA responses, leading to asymmetries between signals recorded from the left and right sides of the body. These differences are particularly evident during specific cognitive and physiological states. Studies have demonstrated that selecting the most relevant side for a given ANS state can improve classification accuracy in human behavior recognition applications, such as cognitive load assessment and sleep stage detection. However, it remains unclear whether incorporating EDA from both sides simultaneously can further enhance performance. In this work, we investigate the impact of leveraging bilateral EDA data on the performance of human behavior recognition models. Specifically, we evaluate its effect on cognitive load/relaxation classification and sleep/wake recognition tasks. Our results indicate that classifiers trained on EDA signals from both sides of the body achieve higher balanced accuracy compared to classifiers trained using unilateral EDA data. These findings highlight the importance of bilateral EDA collection for improving behavior recognition applications.