Sensor-Based Human Activity Recognition for Assistive Health Technologies

Mehr zum Buch

The increase in life expectancy due to advancements in health sciences has led to a growing elderly population, raising concerns about their quality of independent living. Clinicians assess this through Activities of Daily Living (ADLs). With the rise of portable computing devices, a software system that detects ADLs using sensor data from wearable devices can significantly aid in identifying health issues and enhancing healthcare support. This book explores various machine learning approaches for human activity recognition (HAR) using time-series data from wearable sensors in home settings. The first section focuses on machine learning methods for recognizing simple, short-term activities. The second section delves into algorithms for identifying long-term and complex ADLs, introducing a two-stage recognition framework and an online system for continuous HAR monitoring. The final section presents an innovative approach that addresses data scarcity and enhances HAR performance through multitask learning methods, enabling simultaneous training of models for both short- and long-term activities, regardless of their temporal scale. Results indicate that this approach significantly improves classification performance compared to traditional single-task learning methods.