Improved Indoor Positioning Based on Fingerprint Algorithm Using Reference Point Subregion and Access Point Optimization

As indoor positioning becoming increasingly important in the past decade, fingerprint algorithms with high performance are intensively desired. In this paper, an improved fingerprint algorithm with related APs mergence, regional estimation and AP selection is proposed. In the offline phase, the positioning area is evenly divided into a plurality of square subregions and RPs are uniformly distributed at the vertices of the square subregions. After the RSSI vectors at each RP are collected and the relation coefficients between every two APs are calculated, the related APs mergence strategy is used to reduce the storage size of the fingerprint database. In the online phase, after the real-time RSSI vector at the TP was collected, regional estimation and AP selection strategy are used to improve the positioning accuracy. In regional estimation, the TP is estimated to locate in two subregions. In AP selection strategy, the fluctuation ranges and the spatial discrimination abilities of each AP are calculated and the APs whose RSSI values are out of the fluctuation ranges and the APs with poor spatial discrimination abilities are eliminated. Finally, the WKNN algorithm is used to estimate the position. Experiments were performed on three different smartphones to compare the performance of the proposed positioning algorithm with the other three fingerprint algorithms. The experiment results showed that the proposed algorithm reduces the storage size of the fingerprint database by more than 25%. It also reduces the average positioning error by 33.58%, 27.05%, and 13.59% compared with the other three algorithms respectively. The minimum average positioning error of the proposed algorithm is only 0.89 meter.