Performance Evaluation of Radio Fingerprinting Localization

Abstract

This paper introduces a radio fingerprinting localization method for positioning unknown radio transmitters (URTs) based on received signal strength difference (RSSD). The method incorporates Kalman filter (KF) preprocessing, principal component analysis (PCA), similarity measures, and weighted k-nearest neighbors (WKNN). First, the Kalman filter is applied to the received signal strength (RSS) measurements to reduce noise. Next, PCA is used for dimensionality reduction and decorrelation by extracting the principal components from the RSSD data. In the final stage, the similarity between offline and online principal component databases is measured using various metrics, while WKNN estimates the transmitter’s position by assigning weights to nearby reference points (RPs). Simulations are conducted to evaluate the impact of preprocessing, the number of PCA components, and the choice of similarity measures on localization performance. The results provide a comprehensive analysis of the trade-offs between these techniques, highlighting their effectiveness in different environments and conditions for fingerprinting-based WLAN localization.