In recent years, many fingerprint-based localization approaches have been proposed, in which diﬀerent features (e.g., received signal strength (RSS) and channel state information (CSI)) were used as the fingerprints to distinguish diﬀerent positions. Although CSI-based approaches usually achieve higher accuracy than RSSI-based approaches, we find that the localization results of diﬀerent approaches usually com-pensate with each other, and by fusing diﬀerent features we can get more accurate localization results than using only single feature. In this paper, we propose a localization method that fusing diﬀerent features by combining results of diﬀerent localization approaches to achieve higher accuracy. We first select three most possible candidate positions from all the candidate positions generated by diﬀerent approaches according to a newly defined metric called confidence degree, and then use the weighted average of them as the position estimation. When there are more than three candidate positions, we use a minimal-triangle principle to break the tie and select three out of them. Our experiments show that the proposed approach achieves median error of 0.5 m and 1.1 m respec-tively in two typical indoor environments, significantly better than that of approaches using only single feature.
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||12th International Conference on Wireless Algorithms, Systems, and Applications, WASA 2017|
|Period||19/06/17 → 21/06/17|
- Theoretical Computer Science
- Computer Science(all)