This paper presents a novel unsupervised spatial data mining model especially adapted for activity recognition inside a smart home. The goal of our research is to have a scalable, simple to implement model that could enable to recognize the resident's Activities of Daily Living (ADLs). Our algorithm exploits ubiquitous sensors and passive RFID technology to achieve the learning and the recognition. The RFID is used to track all objects in the smart home in real-time. An algorithm then extracts qualitative spatial features from the positions dataset. Finally a clustering is performed with an adapted version of the Flocking algorithm. Our experimental results are very encouraging with a classification rate ranging from 85% to 93%.
Paper
Full text
Unsupervised Spatial Data Mining for Smart Homes
Semantic Scholar · Computer Science · 2015
Abstract
This paper presents a novel unsupervised spatial data mining model especially adapted for activity recognition inside a smart home. The goal of our research is to have a scalable, simple to implement model that could enable to recognize the resident's Activities of Daily Living (ADLs). Our algorithm exploits ubiquitous sensors and passive RFID technology to achieve the learning and the recognition. The RFID is used to track all objects in the smart home in real-time. An algorithm then extracts qualitative spatial features from the positions dataset. Finally a clustering is performed with an adapted version of the Flocking algorithm. Our experimental results are very encouraging with a classification rate ranging from 85% to 93%.