FLUCTUATION CORRECTION FOR K-NEAREST NEIGHBOR LOCATION FINGERPRINTING FOR INDOOR POSITIONING SYSTEM

摘要

MANY INDOOR POSITIONING TECHNIQUES HAVE BEEN PROPOSED AND MANY OF WHICH USE A VARIATION OF FINGERPRINTING METHODS. EVEN SO, ONE OF THE MAIN PROBLEMS THAT LIE WITH FINGERPRINTING IN REAL WORLD APPLICATIONS IS DEALING WITH UNSTABLE WLAN RSS FLUCTUATIONS THAT CAN ADVERSELY AFFECT THE ALGORITHM’S ONLINE PROCESS OF ACCURATELY PREDICTING LOCATIONS CORRECTLY. THE PRESENT INVENTION PROPOSES AN IMPROVED FINGERPRINTING METHOD THAT FIRST MINIMIZES THE FLUCTUATING RSS LEVELS SO THAT A MORE STABLE SIGNAL CAN BE USED FOR THE LOCATION PREDICTION BY THE k-NN ALGORITHM. THE RESULTS PROCURED FROM UTILIZING THIS METHOD OF FILTERING OUT FLUCTUATING RSS LEVELS SHOW BETTER PRECISION (THE PERCENTAGE OF THE NUMBER OF TRIALS OF WHICH THE CORRECT LOCATION IS ESTIMATED OVER THE TOTAL NUMBER OF TRIALS DURING THE EXPERIMENT) WITH THE INCORPORATION OF THE FILTERING ALGORITHM,UP TO 94.38% PRECISION RATE WITH 8 RSS VALUES FROM REFERENCE ACCESS POINTS IN EACH SAMPLE VECTOR IN OUR EXPERIMENTS, COMPARED TO JUST A PRECISION OF ONLY 75.71% WHEN PROCESSING UNFILTERED SAMPLE VECTORS. THE MOST ILLUSTRATIVE FIGURE IS FIGURE 1