Fuzzy support vector machines for device-free localization

Abstract

In this paper, we develop a novel fuzzy support vector machine for device-free localization. The fuzzy support vector machine is an integration of support vector machines (SVMs) and fuzzy systems; therefore a fuzzy system can be extracted from an SVM. We not only show how to integrate SVMs and fuzzy systems, but also show how to reduce the complexity of the obtained fuzzy systems. One major benefit of reducing the complexity of fuzzy systems is that the obtained fuzzy systems are easy to be optimized. The proposed method is proved to be effective through experimental studies, which is carried in a badminton court in which four WiFi access points and 17 test points are deployed. The simulation results show the reduced fuzzy system is easy to perform optimization and generates better results than pure SVM. An simulation result shows the correctness of pure SVM is 66.8% and the correctness of optimized fuzzy systems is 74.6%.