Activity recognition under instances of uncertainty is currently recognised as being a challenge within the application domain of Smart Homes. In such environments, uncertainty can be derived from the reliability of sensor networks, the interaction between human and physical objects and the variability of human activities. Nevertheless, it is a difficult process to quantify these sources of uncertainty within the context of an effective reasoning model in order to accurately recognize activities of daily living (ADL). In this paper we propose an evidential framework, where a sensor network is modelled as an evidence space and a collection of ADLs is subsequently modelled as an activity space. The relation between the two spaces is modelled as a multi-valued probabilistic mapping. We identify two sources of uncertainty in terms of sensor uncertainty and relation uncertainty that is reflected in the interaction between sensors and ADLs. Estimations of the respective types of uncertainty were achieved through manufacture statistics for the sensor technology and by learning statistics from sensor data. A preliminary experimental analysis has been carried out to illustrate the advantage of the proposed approach.
|Title of host publication||Advances in Computational Intelligence Communications in Computer and Information Science|
|Publication status||Published - 2012|