以空間資訊改善房間定位準確度之研究

Using spatial information to improve the accuracy of room-level localization

謝文芝(Wen-Chih Hsieh)

指導教授 ： 項潔

共同指導教授 ： 許永真

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2008年)

https://doi.org/10.6342/NTU.2008.10142

摘要

在室內定位系統中，如果能判斷出被追蹤物的房間位置，我們將可以提供許多有用的服務。然而目前利用訊號強度所做的定位系統準確度大約為兩米左右，若是直接轉為房間資訊，往往會產生錯誤的結果。
本實驗以particle filters機率模型結合likelihood model 與motion model，以Gaussian processes來產生訊號強度的可能性模型(likelihood model)，我們不是在likelihood model中考慮位置對訊號強度造成的影響，我們也加入了方向對訊號強度造成的影響，由於訊號強度與被追蹤物所面對的方向相關，加入了方向的資訊後定位的正確率也因此而提升。另外，並在空間資訊的幫助下我們考慮被追蹤物合理的移動模式讓motion model變得更加合理。在判斷房間位時，我們利用在房間中particle散布的情況和環境中房間的相連情形來決定被追蹤物的房間位置。由於一個人無法劇烈的在不同房間中移動，我們利用likelihood threshold來限制被追蹤物房間位置的改變。
我們以片段錯誤率、序列編輯距離、延遲時間來評估系統的正確性並發現我們提出的方法雖然有大約一秒左右的延遲時間，在片段錯誤率、序列編輯距離的表現上都優於其它方法。

關鍵字

高斯進程 ； 粒子濾波 ； 無線電波訊號強度指標 ； 房間定位

並列摘要

Signal strength-based method is widely adopted in localization nowadays. Because wireless signal strength is unstable, localization deviations are usually larger than one meter. For indoor localization systems, it is essential to provide room-level accuracy. However, with localization deviations larger than one meter, it is difficult to provide accurate room-level location. To solve this problem, we take advantage of spatial information and implement the concept of particle filters. Morever, we apply the Gaussian processes in the likelihood model to predict the mean and variance of the signal strength at any location and direction without the requirement of collecting the wholde training data. Unlike traditional methods of computing likelihood model, which merely consider the influence of location in signal strength, our system takes both location and direction into account. To avoid serious mistakes in localization results, we introduce a threshold in our system to restrict the transition of the tracked person’s room-level location and using spatial information to increase accuracy and efficiency of the constraint. We design a flow to determine whether the system should change the room location in all situations.
To evaluate the results, we compare our system with other methods and apply frame error rate, word error rate and delay time as the performance metric. As a result, within tolerable delay increase, our system performs the best in the frame error rate and word error rate among all localization methods.

並列關鍵字

indoor localization ； particle filters ； RSSI ； room-level ； Gaussian processes

參考文獻

[2] P. Bahl and V. N. Padmanabhan. RADAR: An inbuilding rf-based user location and tracking system. In Proceedings of the Nineteenth Annual Joint Conference of the IEEE Computer and Communications Societies, volume 2, pages 775–784, Tel Aviv, Israel, March 2000.

[3] P. Bahl, N. Padmanabhan Venkata, and A. Balachandran. Enhancements to the radar user location and tracking system. Technical report, Microsoft Research, 2000.

[5] B. Ferris, D. Fox, and N. Lawrence. Wifi-slam using gaussian process latent variable models. In M. M. Veloso, editor, IJCAI, pages 2480–2485, 2007.

[6] B. Ferris, D. Haehnel, and D. Fox. Gaussian processes for signal strength-based location estimation. In Proc. of Robotics: Science and Systems, 2006.

[7] J. Hightower and G. Borriella. Location systems for ubiquitous computing. ”IEEE Computer”, 34(8):57–66, 2001.