Indoor localization by a novel probabilistic approach

The main contribution of our work is to develop a novel localzation algorithm called transformation based probabilistic approach for building a robust and compact localization system. Instead of operating in the measured RSS domain, our algorithm projects the measured signal into a transformed signal space. This approach offers a more efficient mechanism to utilize information of all APs and overcomes the drawback of traditional AP selection techniques which unavoidably throws out the information of unselected APs. Furthermore, the transformation can be viewed as passing through different filters and these filtering processes help reduce the noisy components of the measured signal. Experimental results show that the proposed algorithm demonstrates outstanding performance while the coefficients implementation is based on principal component analysis (PCA) technique. Not only the information compaction, but also the robustness is achieved in our indoor WLAN positioning system. The numerical results show that the error mean is reduced by 42% and the error variance is reduced by 71% on the average.