CSI passive indoor fingerprint positioning method based on improved WKNN

Abstract

Aiming at the problem of low positioning accuracy caused by excessive interference in amplitude and phase construction, a passive indoor positioning method based on channel state information based on improved weighted K nearest neighbor algorithm is proposed. In the offline stage, the isolation forest method, wavelet domain denoising with improved threshold and linear transformation method are used to preprocess the collected channel state information. The processed amplitude and phase information are used as fingerprint data to construct a stable fingerprint database related to the reference point position information. In the online stage, an improved weighted K nearest neighbor algorithm is proposed to repeatedly match the estimated coordinates. After obtaining the position coordinates in one match, the algorithm calculates the Euclidean distance between the K nearest neighbor points of the position coordinates, and uses Gaussian transform to calculate the weights of the K distance values to complete the positioning of the personnel. Experimental simulation tests were carried out in the classroom and the hall respectively. The results show that about 81% of the test position errors of the proposed algorithm are controlled within 1 meter, which can effectively improve the positioning accuracy.