Human Activity Recognition

A field of machine learning and ubiquitous computing that uses sensor data — typically from accelerometers, gyroscopes, and other sensors in smartphones, smartwatches, or other wearable devices — to automatically identify and classify physical activities performed by a person. In accessibility contexts, HAR can enable assistive applications that monitor activities to provide timely support, such as detecting when a student with neurodiversity goes off-task during learning activities or recognizing daily living activities for people with cognitive disabilities. A key challenge is ensuring that HAR models are trained on data from diverse populations including people with disabilities, as models trained solely on neurotypical data may not generalize to users with different motor patterns or behaviours.