Occurrence-Based Fingerprint Clustering for Fast Pattern-Matching Location Determination

Fingerprint clustering is an efficient approach to address the scalability problem of the pattern-matching localization system in large-scale environments. In this letter, we propose an occurrence-based fingerprint clustering algorithm that exploits the useful information embedded in the statistics of received signal strength measurements. Our algorithm allows that a reference point can have more than one fingerprint according to the highest received signal strength, and can be grouped into several clusters according to its highest and second highest received signal strengths. Simulation and experimental results show that our algorithm performs consistently a little better than the peer one without clustering in terms of improved localization accuracy and reduced fingerprint comparisons. Compared with two other clustering algorithms, our algorithm also achieves higher localization accuracy and reduces false cluster selection.