Accuracy enhancement of an indoor ANN-based fingerprinting location system using Kalman filtering

Abstract

This paper presents an accuracy enhancement solution to mobilepsilas location and tracking systems in indoor wireless local area network (WLAN) environments. The enhancement method consists of the Kalman filtering application to an artificial neural network (ANN) based fingerprinting location technique. The application of Kalman filtering has the advantage of using information about the mobilepsilas motion to reduce location errors (caused by the WLAN received signal strength- RSS variations) and to avoid trajectory discontinuities (caused by the static estimation of the ANN-based fingerprinting technique). To process the RSS-based fingeprinting location technique, two ANN-based pattern-matching algorithms have been examined: the generalized regression neural network (GRNN) and the multi-layer perceptron (MLP) and they have been compared to the classic K-nearest neighbors (KNN) method. Experimental results, conducted in a specific in-building environment, showed that the GRNN algorithm performs better than the MLP and KNN algorithms. The application of Kalman filtering to the considered GRNN-based fingerprinting location technique improved the location accuracy of about 22.4 % in terms of location mean error.