Enhancing Multi-Floor Indoor Localization Accuracy Using Fingerprint-Based Dynamic k-NN Approach
Published in 2025 IEEE International Conference on Electro/Information Technology (EIT), 2025
Accurately localizing indoor spaces across multiple floors remains a significant challenge critical for applications such as smart buildings and emergency response systems. This paper presents a novel method for multi-floor indoor localization using Wi-Fi received signal strength (RSS) fingerprinting within a refined k-nearest neighbors (k-NN) framework.
Our contributions are threefold:
Amplified Voting Scheme: A multi-layered strategy for floor differentiation using Euclidean distances, vote counts, and tie-breaking through summed distances. This approach achieved perfect floor classification accuracy in dense datasets and a 2.5% improvement in sparse datasets.
Dynamic Valid Neighbor Selection: Instead of a fixed k, neighbors are selectively added based on RSS similarity thresholds, improving localization precision by excluding outliers and adapting to variable fingerprint density.
Floor-Specific Preprocessing: Each floor’s fingerprint dataset is filtered and normalized independently, reducing the impact of structural and propagation inconsistencies across floors.
Evaluation on a dataset of 832 fingerprints (609 dense, 223 sparse) collected from three floors of an engineering building showed the proposed approach outperforms fixed neighbor methods. Using dynamic neighbors, the mean localization error was reduced to 1.53 meters (dense) and 2.87 meters (sparse), compared to 1.65 meters and 3.04 meters respectively with fixed neighbors.
This framework enables robust, floor-aware localization by adapting to environmental variations and improving spatial accuracy in complex multi-level buildings.
Recommended citation: Benyamain Yacoob, Daniel Marku, and Mina Maleki. 2025. "Enhancing Multi-Floor Indoor Localization Accuracy Using Fingerprint-Based Dynamic k-NN Approach." In Proceedings of the 2025 IEEE International Conference on Electro/Information Technology (EIT). IEEE, USA.
Download Paper | Download Slides