Hierarchical SVM classification for localization in multilevel sensor networks

Abstract

We show that the localization problem for multilevel wireless sensor networks (WSNs) can be solved as a pattern recognition with the use of the Support Vector Machines (SVM) method. In this paper, we propose a novel hierarchical classification method that generalizes the SVM learning and that is based on discriminant functions structured in such a way that it contains the class hierarchy. We study a version of this solution, which uses a hierarchical SVM classifier. We present experimental results the hierarchical SVM classifier for localization in multilevel WSNs. © 2008 Springer-Verlag Berlin Heidelberg.