| Title |
An RF Fingerprint Map Update Technique using Neural Networks |
| Authors |
이성호(Seongho Lee) ; 정성훈(Sunghoon Jung) |
| DOI |
https://doi.org/10.5573/ieie.2021.58.8.3 |
| Keywords |
Indoor localization; Fingerprint; Received signal strength; Deep learning; Dilated convolutional neural network |
| Abstract |
While RF fingerprint-based position estimation has high localization performance, it has the disadvantage of measuring the RF fingerprint map anew whenever the wireless environment changes. To address these challenges, we proposed an RF fingerprint map update technique using neural networks. The proposed technique learns the relationship between indoor structures and wireless environments to predict new fingerprint maps with only information about changes in indoor structures. The proposed technique can predict fingerprint maps for new wireless signal environments, regardless of location, number, size shape, and location of obstacles. The experiment shows localization error is reduced by 0.44m when using a fingerprint map predicted by the proposed technique compared to using a fingerprint map that is not updated. |