Multi-observers instance-based learning approach for indoor symbolic user location determination using IEEE 802.11 signals

Abstract

Wi-Fi's signals strength (SS), and signal quality (SQ) are found to greatly fluctuate in determination of symbolic user location in an indoor environment. This paper explores the influence of several different training data-sets in determining user's symbolic location. The implementation and experimentation were done using off-line instance-based machine learning methods to filter all of the training data-sets. The training data-sets were optimized using “multiple observers” k-Nearest Neighbor approach. Using this method, four different observations were compared, which were 8M observations of SQ and SS , 8M SS observers, 1M SQ and SS and the last was 1M SS observers. Then, a continuing determination of the user location was performed by finding the majority of the nearest ten (k=10) user locations.