Localization with Multidimensional Channel Fingerprints of Multiband Cellular Signals

Precisely predicting the user's location in a Global-Navigation-Satellite-System-degraded environment is challenging. Wireless signal navigation is emerging as a promising solution to this issue and has been garnering increasing attention. Fingerprint localization is well-known for its effectiveness in utilizing multipath or non-line-of-sight (NLOS) signals without requiring precise timing synchronization among transmitter nodes. This paper introduces a novel approach called Multi-Bands and Multi-Dimensions (MBMD) fingerprinting, which integrates frequency information, angle-of-arrival (AOA) values, received signal strength (RSS) values, and time-of-arrival (TOA) values from cellular signals. The angle domain Fourier transform (ADFT) and delay domain Fourier transform (DDFT) is utilized to extract the TOA and AOA information from the estimated channel frequency response. The Subsequently, a weighted K-nearest neighbor (WKNN) algorithm employing Euclidean distance is used for pattern-matching localization based on the MBMD fingerprint. The WKNN algorithm leverages the characteristics of MBMD fingerprints, which encompass dense features directly related to geographic information. Additionally, the WKNN algorithm exhibits strong robustness to environmental variations and diverse fingerprint databases. Experimental results with public dataset demonstrate the superior performance of this proposed approach compared to other fingerprint localization algorithms reported in existing literature. Simultaneously, the experimental results validate that with an increase in the number of dimensions and features in MBMD fingerprints, the accuracy of fingerprint localization improves. keywords: fingerprint localization, cellular signal, TOA, AOA, RSS.