Improved Wknn Indoor Positioning Algorithm Based On C-Means And Chi-Square Distance

摘要

Because of the rapid development of smart cities, WLAN-based location services have also become commonplace. Among the known positioning algorithms, The classical KNN algorithm calculates the Euclidean distance between the undetermined locus and the reference point in the fingerprint database, and choose K points with the smallest Euclidean distance, and takes the arithmetic average of these K points to obtain the predicted value of the undetermined locus; WKNN algorithm is a KNN algorithm improved by weighting, it calculates the predicted value by assigning different weights to the K points. However, such algorithms only consider the absolute distance between RSS vectors at each location. it is common to ignore the relative distance between RSS vectors at various locations. And they can only give each AP the same weight 1/K[1, 2]. In order to overcome the deficiency of the absolute distance of Euclidean distance method, an improved WKNN indoor positioning algorithm based on c-means and chi-square distance was proposed. This method first USES c-means to cluster the fingerprint database, then calculates the weight of each AP with the chi-square distance and sensitivity method, and then corrects the chi-square distance with this weight. The prediction results of the weighted chi-square distance combined with the WKNN treatment registration point show that the accuracy of this method is higher than the traditional WKNN.