Indoor positioning system based on the improved W-KNN algorithm

With the continuous development of intelligent terminal technology, the demand for Location Based Services (LBS) is also increased. Thus, the research of indoor positioning technology is of great significance. Based on the characteristic of the mobile positioning terminal and the presence of the AP in the indoor positioning environment, the indoor location fingerprint algorithm is the most feasible method due to its stable performance without the need for additional hardware. However, the traditional methods do not take the weight of the composition into consideration, which actually affects the precision of positioning much. In this paper, an improved algorithm based on weighted K-nearest neighbors (W-KNN) is proposed. In the off-line stage, the algorithm filters out invalid data by using data expectation, and selects the mean of RSS and the variance of access points (APs) as the eigenvector. In the on-line stage, according to the variance of APs, the weighted distance is proposed to calculate the similarity. Meanwhile, the nearest main neighbor and (K-1) auxiliary neighbors are obtained by comparing the weighted distance. Using the correlation between main neighbor and (K-1) auxiliary neighbors, the location is derived. Experimental results show that the precision of the proposed algorithm is better than that of using the RSSI location fingerprinting algorithm alone in the indoor environment.