標題: 運用三維特徵於智慧型手機上進行人類行為活動辨識
Three Dimensional Features for Human Activity Inference on Smart phones
作者: 賴建翔
Lai, Chien-Hsiang
彭文志
Peng, Wen-Chih
網路工程研究所
關鍵字: 資料探勘;行為;推測;辨識;識別;活動;data mining;activity;inference;identification;recognization
公開日期: 2015
摘要: 挑選最佳的特徵於智慧型手機上進行人類行為活動辨識是ㄧ個富有挑戰性的課題。在本論文中,我們運用三維資料來解決於智慧型手機上進行活動辨識所遇到的困難: 第一,時空資料: 反映日常慣例以及地理位置的語意。第二,應用程式: 感知專門協助使用者進行行為活動的應用程式。第三,動作: 用於識別運動類型的行為活動。除了智慧型手機資料之外,我們也利用外部資料集提出公眾意見來推測使用者的行為活動。最後,我們在真實資料集上進行實驗並與常見的分類方法進行比較來評估我們提出的方法的有效性和效能。實驗結果顯示我們的方法優於其他的分類方法無論是在準確度、執行時間或是模型儲存效率方面。
Selecting best features for human activity inference in smart phone is a challenging task. In this paper, we tackle the problem of activity inference in smart phone by utilizing three dimensions of smart phone data: 1) Spatial-Temporal data: reflecting daily routines and semantic of locations, 2) Application: perceiving specialized apps that assist user’s activities, 3) Motion: distinguishing motion-type activities from others. In addition to smart phone data, we utilize outsourced dataset to address public opinions to infer users’ activities. Finally, we compare our proposed method with several common classification methods on real dataset to evaluate the effectiveness and performance of our method. Experimental results show that our approach outperformed other methods in terms of accuracy, running time, and storage efficiency.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT070256509
http://hdl.handle.net/11536/127348
Appears in Collections:Thesis