A Novel and Distributed Approach for Activity Recognition Inside Smart Homes

The human life expectancy has never stopped increasing during the last decades causing, among other benefic phenomenon, the apparition of different types of dementia often due to the Alzheimer disease. Despite a reduced independence these patients want to stay at their home more than anything causing the apparition of an expensive home care. To answer this problem, smart homes appeared, and many works focused on recognize the activities made by the inhabitant to assist him if the need arises. However, only a few authors worked on creating a cheap and reliable smart home infrastructure. We already addressed this point in previous works by creating a distributed architecture. Still, we did not prove, at the time, that the activity recognition was still achievable on such infrastructure. Consequently, we present, in this paper, a new kind of distributed activity recognition implementing a form of safety based on reflex behavior and making all its decision in a distributed manner. With an average accuracy at least as good as the centralized manner, we prove here that even if the architecture is distributed, the activities can still be recognized.