Research on Location Fingerprint Based WiFi Positioning Algorithm

10.3966/199115992017102805015

Maoyuan Huang；Shaobo Wu；Yinglei Li

clustering ； fingerprint ； k-means ； positioning algorithm ； RSSI ； WiFi ； WKNN ； WLAN

電腦學刊

28卷5期（2017 / 10 / 01）

162 - 177

英文

Existing WiFi signals are utilized to carry out indoor positioning, and received signal strength RSSI is usually selected as a positioning feature parameter. An RSSI positioning algorithm is divided into a range-based positioning algorithm and a range-free positioning algorithm. The range-based positioning algorithm calculates a distance by utilizing an indoor transmission loss model and is lager in dependency on an indoor model; and the range-free positioning algorithm adopts a location fingerprint algorithm, a fingerprint database is established by only needing to measure RSSI values, each fingerprint is in only correspondence to one location information, information of an unknown location can be estimated by matching the unknown location fingerprint and the fingerprint database, and realization is simple. This article studies the positioning principle of the location fingerprint algorithm, indicates influence factors which may generate errors in the positioning process, comprehensively and deeply analyzes sources of the errors and also indicates the limitation of the existing location estimation algorithm. This article presents a k-means and WKNN based location fingerprint algorithm. According to the location fingerprint algorithm, the fingerprint database is preliminarily established by measuring collected RSSI values for many times and solving a mean and then is trained by utilizing k-means clustering analysis, and some fingerprints with very small similarity are removed; and actually measured fingerprints are matched with the fingerprint database after being trained, so that the accuracy of the fingerprint database is improved, the search space of matching is reduced, and the influence of the fingerprint database on a positioning result is reduced. In the location estimation stage, a new weight coefficient calculation method is introduced, the location accuracy of WKNN is improved.