傅立成Fu, Li-Chen臺灣大學:資訊工程學研究所何育誠Ho, Yu-ChenYu-ChenHo2010-06-022018-07-052010-06-022018-07-052008U0001-1508200821331500http://ntur.lib.ntu.edu.tw//handle/246246/184958本論文提出在智慧環境中學習及便是人類日常生活的問題，先前大部分的方法先收集人類行為的資料並學出其模型，然後再使用學得的模型來辨識人的行為，然而，人的行為習慣及環境的佈置可能會隨著時間而發生改變，造成行為的模式發生改變，這時舊有的辨識用的行為模型便過時了，使辨識率降低，必須要重新學習新的行為模型，但是重新收集學習用的行為資料並給予對應的行為標籤是件非常煩人且容易出錯的工作，在這樣的情況下，在更新行為模型時能降低人為的指導工作份量是件非常重要的事，本論文提出一個可以自我調整行為模型的行為辨識方法，它可以在動態的環境下同時辨識多種行為，並以較少的人為指導來跟著環境變動調整行為模型。This thesis addresses the problem of learning and recognizing human daily activities in smart environment. Most approaches offline learn the activity model and recognize the activity in an online phase. However, the activity models can be outdated when the human behavior and environment deployment change. It is a tedious and error-prone job to recollect data for retraining the activity models. In such case, it is important to adapt the learnt activity models under one context to another context without too much supervision. In this thesis, we present a self-reconfigurable approach for activity recognition can reconfigure a previously learned activity model to infer multiple activities under a dynamic environment meanwhile requiring minimal human supervision for labeling training data.誌謝 i文摘要 iiBSTRACT iiiONTENTS ivIST OF FIGURES viiIST OF TABLES xhapter 1 Introduction 1.1 Motivation 1.2 Challenges of Activity Recognition 3.3 Objectives 4.4 Related Work 6.5 Thesis Organization 8hapter 2 Preliminaries 10.1 Problem Statement 10.2 System Overview 11.3 Dynamic Bayesian Networks (DBNs) 14.3.1 DBNs: Representation 15.3.2 Inference in DBNs 17.3.3 DBNs: Learning 19.4 Sufficient Statistics 20.5 Semi-supervised Learning 20.5.1 Expectation-Maximization (EM) algorithm 21.6 Active Learning 22.7 Online Learning 23hapter 3 Activity Recognition System in a Static Environment 24.1 Overview 24.2 Environment Sensors and Interaction Detectors 27.2.1 Sensor Deployment 27.2.2 Interaction Detectors 30.3 Activity Modeling 32.3.1 Feature Generation 36.3.2 Activity Model 38.4 Model Learning 41.4.1 Feature Selection 42.4.2 Parameter Estimation 43.4.3 Used Sufficient Statistics in Learning Procedure 44.5 Activity Recognition 44hapter 4 Activity Recognition System in a Dynamic Environment 46.1 Overview 46.2 Self-reconfiguring 50.3 Active Learning for Activity Label Requirement 53hapter 5 System Evaluation 55.1 Experiment Environment 55.2 Evaluation Description 58.3 Evaluation Metric 59.4 Experimental Result and Discussion 62.5 Fall Detection Application 67hapter 6 Conclusion 69.1 Summary 69.2 Future Work 70.2.1 Improving Environment Sensors 70.2.2 Reducing the Learning Effort 71.2.3 Improving the Self-reconfigurable Activity Recognition System 71EFERENCE 73application/pdf2381189 bytesapplication/pdfen-US行為辨識機率推論動態貝氏網路主動式學習Activity RecognitionProbabilistic ReasoningDynamic Bayesian NetworkActive Learningthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/184958/1/ntu-97-R95922026-1.pdf