A Grammar Inference Algorithm for Event Recognition in Sensor Networks
Sahin Cem Geyik, Boleslaw Karol Szymanski · 2008
Abstract—In this paper, we demonstrate that probabilistic context free grammars (PCFGs) can be used to recognize events from a sensor data stream. A fast PCFG inference algorithm based on Stolcke(1994) and Chen(1996) is presented which utilizes the set of observation strings as training data. A realworld scenario is presented and we also show that multi-modal sensor information can be utilized using Dempster-Shafer theory of evidence. I. INTRODUCTION AND PREVIOUS WORK Today’s sensor applications collect vast amount of measurement data, so it is increasingly important to provide their users with high level representation of those measurements. The focus of this paper is on using Probabilistic Context