MFMCF: A Novel Indoor Location Method Combining Multiple Fingerprints and Multiple Classifiers

Abstract

WiFi fingerprint-based localization has attracted significant research interest recently because WiFi devices were widely deployed and practicable. The accuracy of indoor positioning based on single fingerprint pattern is limited since it is susceptible to external influences. This paper proposes a multi-fingerprint and multi-classifier fusion(MFMCF) localization method, which improves the localization accuracy by constructing multi-pattern fingerprints and integrating multiple classifier. MFMCF constructs signal strength difference(SSD), hyperbolic location fingerprint(HLF) and received signal strength(RSS) as a composite fingerprint set(CFS) using linear discriminant analysis(LDA). A special decision-structure of multiple classification was designed by calculating the entropy of the classifiers including K-Nearest Neighbor(KNN), Support Vector Ma-chine(SVM) and Random Forest(RF), to obtain a more accurate estimate result. Experiments show that MFMCF has higher localization accuracy and robustness to single fingerprinting pattern.