使用智慧型手機上加速器辨識交通工具

Transportation Mode Detection with Single Accelerometer on Smart Phones

張育誠(Yu-Chen Chang)

指導教授 ： 薛智文

共同指導教授 ： 許永真(Yung-Jen Hsu)

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2010年)

https://doi.org/10.6342/NTU.2010.00291

摘要

從嵌入於攜帶型日常生活物品的感應器中，得知使用者移動行為在普及計算領域中是一個顯學。作為一種人類活動，交通模式，例如走路、騎腳踏車、騎機車、開車等，可以提供知識，用以幫助了解移動行為。這篇碩士論文探索如何用單一、低性能、裝置於手機上的加速器資料來辨識高複雜度使用者交通模式的屬性。我們的方法同時考慮了一般對於交通工具所在基礎建設的常識，以及一般使用者攜帶手機的習慣。藉由消除角度特型和重新標記的技術，我們建立了因使用者行為和交通工具移動導致的震動這層資料，然後從中擷取可辨識的特徵，來進行交通模式辨認。在十七個使用者在至少一個月的日常生活中，從標記的八百三十一筆移動軌跡驗證，整體而言，在辨識都市地區常見的六種交通模式，我們的系統在以整趟移動為樣本的事後分析獲得平均89%的準確度，在以時間視窗為樣本的即時偵測中獲得平均78%的準確度。

關鍵字

手機 ； 加速器 ； 情境感知 ； 行為辨識 ； 交通模式

並列摘要

Learning user mobility from sensor embedded in portable everyday object is a dominant research area in pervasive computing. As a kind of human activity, transportation modes,such as walking, cycling, riding, driving, and etc., can provide more knowledge for mobility understanding. This thesis explores how single low-level acceleromter data from smart phones can be used to recognize high-level properties of user transportation. Our method considers both commonsense constraint of transportation infrastructure and regular user behavior on carrying mobile phone. With de-orientation and relabeling, we constructed the vibration, which was caused by both user action and vehicle motion, layer and extracted discriminable pattern from it for transportation mode inference. Evaluated with 831 user-labeled trails from the daily lives of 17 data collectors over a period of one month, our system got an overall average accuracy of 89% for trail-based analysis and 78% for window-based analysis on 6 kinds of transportation in urban city.

並列關鍵字

Mobile Phone ； Accelerometer ； Context Sensing ； Activity Recognition ； Transportation Mode

參考文獻

[3] D. Ashbrook and T. Starner. Using gps to learn significant locations and predict movement across multiple users. Personal Ubiquitous Comput., 7(5):275–286, 2003.

[4] L. Bao and S. S. Intille. Activity recognition from user-annotated acceleration data. Pervasive 2004, pages 1–17, April 2004.

[5] C. V. Bouten, K. T. Koekkoek, M. Verduin, R. Kodde, and J. D. Janssen. A triaxial accelerometer and portable data processing unit for the assessment of daily physical activity. IEEE Trans Biomed Eng, 44(3):136–147, March 1997.

[6] T. Brezmes, J.-L. Gorricho, and J. Cotrina. Activity recognition from accelerometer data on a mobile phone. In IWANN ’09: Proceedings of the 10th International Work-Conference on Artificial Neural Networks, pages 796–799, Berlin, Heidelberg, 2009. Springer-Verlag.

[7] S.-H. Cha and S. N. Srihari. On measuring the distance between histograms. Pattern Recognition, 35(6):1355–1370, June 2002.