Deep Human Action Recognition System for Assistance of Alzheimer’s Patients

People with Alzheimer’s disease find it difficult to perform their Activities of Daily Living (ADLs) due to limited abilities in cognitive functioning. They cannot remember the correct sequence of the performed activity steps, and sometimes they can spend a lot of time in an activity without finishing it. Therefore, to be successful with ADLs, they need human caregivers, but this reduces their autonomy. In response to these needs of the elderly, we have developed an assistance system for people with Alzheimer. The assistance system aims to ensure ADLs (like drinking beverage) performing correctly, based on action recognition from a video. Our assistance system provides patient assistance prompts as needed. The DemCare dataset was used to develop and validate the proposed assistance system. Obtained results are very promising and are comparable to the state-of-the-art approaches ones.