Summary
International Conference on Emerging Technologies for Communications
2020
Session Number:P1
Session:
Number:P1-4
Training Data Generation for RSSI-based Localization with Camera Object Detection
Tomoya Sunami, Takayuki Nishio, Masahiro Morikura, Koji Yamamoto,
pp.-
Publication Date:2020/12/2
Online ISSN:2188-5079
DOI:10.34385/proc.63.P1-4
PDF download
Summary:
Received signal strength indicator (RSSI)-based indoor localization, which uses the radio frequency signals of Wi-Fi and Bluetooth devices, has attracted significant attention for the expansion of location-based services owning to its simplicity. In the localization, RSSI values are utilized as a fingerprint to identify the location, and a mapping from the fingerprint to its location is learned through supervised learning. However, to obtain an accurate localization model, it is necessary to generate a large number of labeled training data, which is costly. This paper proposes an automatic data generation method for RSSI-based localization. The proposed method utilizes a camera and object detection for annotating the measured RSSI values. The experimental evaluation demonstrates that the proposed method can generate data samples automatically and improve the prediction accuracy by enabling the training of large-capacity models.