2023/04/30 by Yasith Amarasinghe, Darshana Sandaruwan, Amarasinghe, Yasith +7
Computer Science · Engineering · Health Professions · #Context-Aware Activity Recognition Systems #FOS: Computer and information sciences #Green IT and Sustainability #Human-Computer Interaction (cs.HC) #Mobile Health and mHealth Applications
paper · pdf · doi:10.48550/arxiv.2305.00517
openalex publication_date 2023/04/30 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Energy Expenditure Estimation (EEE) is vital for maintaining weight, managing chronic diseases, achieving fitness goals, and improving overall health and well-being. Gold standard measurements for energy expenditure are expensive and time-consuming, hence limiting utility and adoption. Prior work has used wearable sensors for EEE as a workaround. Moreover, earables (ear-worn sensing devices such as earbuds) have recently emerged as a sub-category of wearables with unique characteristics (i.e., small form factor, high adoption) and positioning on the human body (i.e., robust to motion, high stability, facing thin skin), opening up a novel sensing opportunity. However, earables with multimodal sensors have rarely been used for EEE, with data collected in multiple activity types. Further, it is unknown how earable sensors perform compared to standard wearable sensors worn on other body positions. In this study, using a publicly available dataset gathered from 17 participants, we evaluate the EEE performance using multimodal sensors of earable devices to show that an MAE of 0.5 MET (RMSE = 0.67) can be achieved. Furthermore, we compare the EEE performance of three commercial wearable devices with the earable, demonstrating competitive performance of earables