Accuracy Enhancement for Fingerprint-Based WLAN Indoor Probability Positioning Algorithm

This paper proposed an accuracy enhancement for fingerprint-based WLAN indoor probability positioning algorithm. Based on Bayes´ probability theory, the traditional probability positioning algorithm converts the new sample location coordinates corresponding to the posterior probability to a priori marginal probability, which leads to the reference point having the greatest product for the estimated location as the terminal coordinates. However, due to the limited quantity of recorded signal strength in the offline phase, new recorded signal strength, different from any signal strengths in the Radio map, will appear in the on-line phase, which means the reference point may not completely characterize the signal strength distribution of this point and result in a poorer poisoning accuracy Therefore, based on the accuracy of enhanced probabilistic location algorithm, this paper proposes to use Gaussian and polynomial continuous curve to realize the least square fitting for the original signal discrete intensity distribution, which achieves to improve the positioning accuracy. And also, this paper presents an experiment made in a WLAN indoor open environment with 8×9m², and the results verify the feasibility and validity of the proposed algorithm.