Location fingerprint clustering based on fuzzy kernel c-means algorithm

A fuzzy kernel c-means clustering algorithm(FKC) is proposed to resolve the location fingerprint(LF) clustering. LF is summarized as a kind of interval-valued data which obey normal distribution to describe sampling uncertainty of received signal strength of access point. After mapping LF into the high-dimensional feature space through normal distribution function determined by interval median and size, LF is clustered with fuzzy c-means algorithm based on kernel method in the feature space. Results of ZigBee positioning experiments show that FKC can get better clustering effect than c-means algorithm based on the average value of signal strength. On the premise of ensuring the positioning precision, a feasible solution is provided to decrease the positioning calculation consumption remarkably.