This paper explores a revised evidential lattice structure designed for the purposes of activity recognition within Smart Homes. The proposed structure consists of three layers, an object layer, a context layer and an activity layer. These layers can be used to combine the mass functions derived from sensors along with sensor context and can subsequently be used to infer activities. We present the details of configuring the activity recognition process and perform an analysis on the relationship between the number of sensors and the number of layers. We also present the details of an empirical study on two public data sets. The results from this work has demonstrated that the proposed method is capable of correctly detecting activities with a high degree of accuracy (84.27%) with a dataset from MIT  and 82.49% with a dataset from the University of Amsterdam.
|Title of host publication||Knowledge Science, Engineering and Management Lecture Notes in Computer Science|
|Publication status||Published (in print/issue) - 2010|