Viusal fall alert service in low computational power device to assist persons´ with dementia

In this paper, we propose a computationally efficient fall detection algorithm which exploits visual observations. The proposed architecture is designed to be suitable for devices of low processing capabilities allowing a large scale implementation of the proposed IT technology in the area of aiding elderly or persons´ with dementia. In contrast to other approaches, visual surveillance is more natural way for detecting persons´ falls, especially for people of dementia where falls are often repeatedly. The algorithm has been evaluated in the lab and real-life clinical conditions, indicating a robust fall detection even in cases of high environmental changes.