An Enhanced Subject-Independent Approach for Hand-Based Micro Activities Recognition

Human activity recognition (HAR) is crucial in various fields, including healthcare, assistive technologies, and human-computer interaction. Recognizing hand-based micro activities presents a unique challenge due to their complexity and variability. In this paper, we propose a novel HAR methodology to identify 24 distinct hand-based Activities of Daily Living (ADLs) using data from a wrist-weared inertial sensor. Our approach, tested on a dataset of 30 participants, employs a unique two- level segmentation and lightweight machine-learning model to handle the complex dynamics of micro-ADLs. The experimental outcomes, showcasing an overall accuracy of 84% and precision, recall, and F1-Scores above 83% underscore the method's efficacy and subject-independent reliability compared to the existing literature on micro-ADL recognition.