KNN/PFCM HYBRID MEHOD USING GUSTAFSON-KESSEKL METHOD FOR INDOOR LOCATION DETERMINATION IN WALN

摘要

PURPOSE: A PFCM(Possibilistic Fuzzy C-Means) grouping based mixed wireless indoor positioning method which uses KNN(K-Nearest Neighbor) and GK(Gustafson-Kessel) methods is provided to measure a signal noise ratio received from a plurality of access points in a training step, thereby building a wireless fingerprint type database. CONSTITUTION: A database is built with first propagation characteristic value data in a wireless fingerprint mode(S110). A standard point is searched using a PFCM(Possibilistic Fuzzy C-Means) mixed technique. k standard points are selected using a KNN(K-Nearest Neighbor) technique(S120). A group of the first propagation characteristic value data of the k standard points is formed(S130). A group which has a minimum GK(Gustafson-Kessel) distance with respect to the first propagation characteristic value data and a group center vector are selected(S140). The average of standard point position coordinates belonged to the selected group is calculated. The position of a terminal is estimated(S150).