結合無線訊號強度與攜帶模式辨識資訊之深度學習室內定位方法

A Deep Learning-based Indoor-Positioning Approach using RSSI and Carrying Mode Information

石銘謙(Ming-Chien Shih)

指導教授 ： 林斯寅

中原大學/商學院/資訊管理研究所/碩士(2019年)

https://doi.org/10.6840/cycu201900066

摘要

由於近年來智慧型手機的普及化，以及智慧商務的發展等因素，室內定位為備受矚目的研究主題。以定位基礎的服務系統（Location-based service, LBS）皆可看到定位技術的應用。一個典型的室內定位系統是藉由分析特定無線訊號的強度來作為定位的依據，利用分析Wi-Fi或Bluetooth的訊號強度以三角訂定位的形式取得使用者於空間中的定位。但在當前的定位系統中，主流採用的無線訊號容易在室內的環境受到干擾，使得系統的定位表象不理想。本研究聚焦於討論攜帶模式（Carrying mode）對於定位系統的影響，並以智慧型手機內部的動態感測器（inertial motion sensor）分析使用者當前的攜帶模式。進而提出一個以深度學習模型分析無線強度資訊與攜帶模式資訊（Carrying mode information）的定位系統。本實驗討論了不同種機器學習模型對於定為結果的影響，其中以卷積神經網路可以達到96％的精準度。本實驗證實了攜帶模式資訊有助於增進室內定位的精確表現。

關鍵字

攜帶模式資訊 ； 室內定位 ； 深度學習 ； 模式識別

並列摘要

Indoor smartphone positioning is an enabling technology used to create new opportunity in indoor navigation and mobile location-based services (LBS) applications. In a general scheme, implementing an indoor positioning system relies on the technology of wireless sensor network. The typical solutions are using wireless signals such as WiFi and Bluetooth to construct a system. However, the wireless signals could be influenced by uncertain indoor objects, and the user’s carrying mode of a smartphone. Therefore, how the carrying mode information influencing a positioning system is discussed in the paper. Moreover, we propose an approach to identify the users’ carrying mode of a smartphone, by smartphone’s inertial sensors, to assist an indoor positioning. The deep learning algorithms are deployed in our proposed system. The result represents our system can reach 99% positioning accuracy. In the meantime, carrying mode information is validated that it is able to improve a positioning system.

並列關鍵字

Carrying mode information ； Indoor positioning ； Deep learning ； Pattern recognition

參考文獻

1. Wang, J., Ghosh, R.K., Das, S.K.: A survey on sensor localization. J. Control Theory Appl. 8, 2–11 (2010).

2. He, S., Chan, S.H.G.: Wi-Fi Fingerprint-Based Indoor Positioning: Recent Advances and Comparisons. IEEE Commun. Surv. Tutor. 18, 466–490 (2016).

3. Wang, X., Gao, L., Mao, S., Pandey, S.: CSI-Based Fingerprinting for Indoor Localization: A Deep Learning Approach. IEEE Trans. Veh. Technol. 66, 763–776 (2017).

4. Faragher, R., Harle, R.: An analysis of the accuracy of bluetooth low energy for indoor positioning applications. 1, 201–210 (2014).

5. Gan, X., Yu, B., Huang, L., Li, Y.: Deep learning for weights training and indoor positioning using multi-sensor fingerprint. In: 2017 International Conference on Indoor Positioning and Indoor Navigation (IPIN). pp. 1–7 (2017).