在這個研究中，我們假設老人是獨自生活在一個環境裡，我們利用RFID的技術來偵測老人是否有異常的行為。RFID技術的基本架構是由Reader去讀取Tag，而Reader在讀取Tags訊號時，因為訊號強弱以及距離遠近的關係，我們可以獲得Tag訊號的強弱值(RSSI值 0~255)。
    我們希望以RSSI值的數據來判斷和記錄長輩平日在家移動的數據資料，我們計畫的RFID系統架構是我們在生活的環境中佈滿許多
Tags，例如客廳角落放置、餐廳、廁所、臥房……等等，而Reader則由老人隨身攜帶，跟隨著老人日常生活習慣去移動，亦即時記錄所讀取到的RSSI值，在藉由機器學習的技術來建立一個“正常”行為的模型，進而用來判斷其後續是否屬於異常，若是異常，則發出警訊。
    這樣的方法，不同於許多其他的以電腦視覺為主的方法，不僅僅是偵測老人是否跌倒而已，藉由平常紀錄的RSSI值，可以找到老人的各種生活模式，除了跌倒，是否有午睡習慣、上廁所頻率、開冰箱的
紀錄……等等。我們都可以利用機器學習的方法學習並判斷。另外，我們利用RSSI值的這個方法，並不像一些偵測跌倒異常的偵測技術必須要利用到攝影機，更是減低了一般人對隱私權的疑慮。
    而最後的實驗結果中，有幾筆測試資料的結果非常好，也足以證明我們利用群集分析來判斷異常行為這樣的想法與理論的確是可以行的通，雖然我們並無法100%可以將異常生活行為模式的資料點給找出來，但我們所做的群集分析結果，對於一個老人的行為分析，確實可以提供相當程度的參考價值。在未來的研究方向中，或許我們可以再使用許多更精密的群集分析技術來提升我們偵測異常行為的準確率。

In this research, we aim at building an intelligent system that can detect abnormal behavior for the home-alone elderly at home. Deployment of RFID tags at home helps us collect the daily movement data of the elderly. The RFID technology uses a reader to detect tags. When the reader detects the signals from the tags, RSSI values (0-255) that represent signal strength can be obtained.
    We want to build a behavior model (viewed as normal) for an elderly person by a set of long-term collected RSSI data. We then can use this model to detect subsequent “abnormal” behavior of the person. The system architecture includes deploying active tags in the living environment, e.g., the living room, the dining room, the kitchen, the rest room, and the bed rooms. The reader is to be carried by an elderly person. The detected RSSI values are recorded following the movement of the person. The behavior model built by machine learning can be used to determine if the subsequent behavior is normal or not. Since only data of “normal” behavior are collected, the model (a classifier) can be build by only positive examples.
    This approach, different from computer-vision-based approaches, not just detects predefined events like if the elderly person falls, but finds the living patterns of the person. The machine learning technique can be used to learn all the patterns without defining all of them in advance. Also, it is not necessary to install cameras at home. This can relieve the concern of personal privacy issues. Researches related to using RFID or other sensors for elderly care can also be found, but all of them do not focus on behavior modeling as we do.
    In the experiments, several test results are very successful. This is enough to prove that utilizing clustering analysis here is really practical. Though our system cannot perfectly detect abnormal behavior, it certainly provides valuable information to the analysis of human behavior at home. In the future, using more sophisticated clustering techniques should to able to improve the detection of abnormal behavior.

第一章 緒論................................................1
1.1 研究背景...............................................1
1.2 研究動機...............................................3
1.3 論文組織與架構.........................................5
第二章 文獻分析............................................7
2.1 什麼是RFID？...........................................7
2.1.1 RFID原理與組成.......................................7
2.1.2 RFID與 Bar Code之比較...............................12
2.2異常偵測...............................................14
2.2.1國內外異常偵測相關研究...............................16
2.3 RFID於醫療或照護之應用與研究..........................18
2.3.1 國內外RFID於醫療或居家照護相關研究..................18
2.3.2 RFID於醫療管理之應用................................19
2.4 群集分析 (CLUSTER ANALYSIS)...........................21
2.4.1相似度的計算與測量...................................23
2.4.2 分群法的採用........................................26
2.4.3 K-means分群法.......................................27
第三章 行為模式之建立.....................................30
3.1設備與初步測試環境簡介.................................30
3.1.1現有設備.............................................30
3.1.2 初步測試環境........................................31
3.2資料收集以及前置處理...................................32
3.2.1 處理驟降為0的情形...................................38
3.2.2 Smoothing...........................................41
3.2.3 Sampling............................................43
3.3 機器學習於行為模式之建立..............................44
第四章 系統架構與實驗結果.................................53
4.1實驗環境與行為模式.....................................53
4.1.1 正常行為模式........................................53
4.1.2 異常行為資料........................................56
4.2系統架構與流程.........................................57
4.3不同前置處理方式的影響.................................62
4.3.1 單筆資料不同時間長度的差異..........................65
4.3.2不同時間長度處理訊號干擾範圍的差異...................68
4.3.3 不同Smoothing範圍的差異.............................71
4.3.4不同Sampling範圍的差異...............................71
4.4定義異常警報...........................................72
4.5實驗結果...............................................74
4.5.1時間點計算公式.......................................74
4.5.2 Short Term 短時間異常...............................75
4.5.3 Long Term 長時間異常................................92
第五章 結論與未來展望.....................................96
參考文獻..................................................98
附錄 英文論文............................................104

圖目錄
圖1、RFID 架構示意圖	8
圖2、門框型讀取器	11           
圖3、卡片型讀取器	11
圖4、群集分析步驟	22
圖5、兩種距離在二維空間上的物理意義	24
圖6、群集間相似程度	25
圖7、K-means分群法流程圖	29
圖8、筆記型電腦	30          
圖9、無線網路分享器	30
圖10、CF卡介面PDA	30      
圖11、CF卡 RFID Reader	30
圖12、初步測試環境平面圖	32
圖13、系統架構圖	33
圖14、測試移動路線圖	34
圖15、客廳小桌RSSI值曲線圖	34
圖16、茶桌RSSI值曲線圖	35
圖17、餐桌B RSSI值曲線圖	35
圖18、廁所RSSI值曲線圖	36
圖19、手環RSSI值曲線圖	36
圖20、RSSI值受干擾之現象	37
圖21、RSSI值訊號受到干擾之情形	39
圖22、相加後取平均值	40
圖23、驟降為0情形處理前客廳小桌RSSI值曲線圖	40
圖24、驟降為0情形處理後客廳小桌RSSI值曲線圖	41
圖25、Tags距離示意圖	43
圖26、RSSI值Sampling示意圖	44
圖27、原始資料格式	45
圖28、轉換後RSSI值格式	45
圖29、以五秒鐘為一筆資料點做Sampling	46
圖30、資料時間維度轉換	47
圖31、輸入群集分析資料格式	48
圖32、各點所屬的群集	49
圖33、中心點的座標與其D_MAX	50
圖34、群集間可能相互包含	51
圖35、群集分析使用示意圖	52
圖36、測試環境平面圖	54
圖37、測試環境系統架構圖	58
圖38、Training正常行為模式資料流程圖	60
圖39、Testing異常資料流程圖	61
圖40、資料前處理檔案名稱說明	64
圖41、停留在走道位置圖	88
圖42、第六筆測試資料結果分析圖	92

表目錄
表1、97年中華民國人口年齡分配統計表	2
表2、RFID標籤型式一覽表	9
表3、主動式與被動式RFID標籤比較表	10
表4、RFID讀取器分類表	11
表5、RFID 頻率規格	13
表6、Bar Code與 RFID之比較	14
表7、分群法類型	27
表8、主動式RFID Tag類型及規格	31
表9、SYRD245-CF1讀卡程式欄位資訊說明	45
表10、1小時正常生活行為模式	55
表11、8小時正常生活行為模式	55
表12、30sec_Ns11sec_Sm11se_Sa5sec群集分析結果	67
表13、60sec_Ns11sec_Sm11se_Sa5sec群集分析結果	67
表14、300sec_Ns11sec_Sm11se_Sa5sec群集分析結果	68
表15、30sec_Ns21sec_Sm11se_Sa5sec群集分析結果	70
表16、第一筆測試資料群集分析結果I	76
表17、第一筆測試資料於廁所群集資料點I	76
表18、第二筆測試資料Short Term群集分析結果I	77
表19、第二筆測試資料於餐廳群集資料點I	77
表20、第三筆測試資料群集分析結果I	79
表21、第四筆測試資料群集分析結果I	81
表22、第四筆測試資料於廁所群集資料點I	82
表23、第四筆測試資料以單筆資料5分鐘群集分析結果I	82
表24、第五筆測試資料群集分析結果I	83
表25、第六筆測試資料群集分析結果I	84
表26、第一筆測試資料群集分析結果II	85
表27、第二筆測試資料群集分析結果II	86
表28、第二筆測試資料於廚房群集分析結果II	86
表29、第三筆測試資料群集分析結果II	87
表30、第三筆測試資料誤判於Room1群集的異常資料點	88
表31、第四筆測試資料群集分析結果II	89
表32、第四筆測試資料於廁所群集分析結果II	89
表33、第五筆測試資料群集分析結果II	90
表34、第六筆測試資料群集分析結果II	91
表35、Long Term Behavior Testing Data	93
表36、Long Term Testing Data 群集分析結果	95

[1] Xiao, Y., Shen, X., Sun, B., Cai, L.: Security and Privacy in RFID and Applications in Telemedicine. In: IEEE Communications Magazine, vol. 44, Issue 4, pp. 64--72 (2006)

[2] Lin, C.P., Tai, J.C., Song, K.T.: Traffic Monitoring based on Real-Time Image Tracking. In:  2003 IEEE International Conference on Robotics and Automation, vol. 2, pp. 2091-- 2096 (2003).

[3] S. T. Tseng, K.T. Song, : Real-time image tracking for traffic monitoring, In: Intelligent Transport Systems IEEE International Conference, pp. 1-6. (2002)

[4] Prasithsangaree, P., Krishnamurthy, P., Chrysanthis, P.K.: On Indoor Position Location with Wireless LANs. In: 13th IEEE Int’l Symp. On Personal, Indoor, & Mobile Radio Communications, vol. 2, pp. 720—724, Lisbon, Portugal (2002).

[5] Liau, W.H.: Inhabitant Tracking and Service Provision in an Intelligent e-Home via Floor Load Sensors. Master’s Thesis, Department of Computer Science and Information Engineering, Nation Taiwan University, Taiwan (2005)

[6] Hara, K., Omori, T, Ueno, R.: Detection of Unusual Human Behavior in Intelligent House. In: 12th IEEE Workshop on Neural Networks for Signal Processing, pp. 697--706 (2002)

[7] Chung, P.-C., Liu, C.-D.: A Daily Behavior Enabled Hidden Markov Model for Human Behavior Understanding. In: Pattern Recognition, vol. 40, pp. 1572--1580, Elsevier (2008)

[8] Lee, B., Kim, H.: A Design of Context aware Smart Home Safety Management Using Networked RFID and Sensor. In: First IFIP WG6.2 Home Networking Conference, Paris, France, pp. 215--224 (2007)

[9] Jih, W.-R., Hsu, J.Y., Tsai, T.-M., Context-Aware Service Integration for Elderly Care in a Smart Environment. In: 2006 AAAI Workshop on Modeling and Retrieval of Context Retrieval of Context, pp. 44—48, The AAAI Press, Menlo Park, California (2006)

[10] RFID Journal, http://www.rfidjournal.com

[11] Crepaldi, R,, Harris, A.F., Kooper, R., Kravets, R., Maselli, G., Petrioli, C., Zorzi,  M.:  Managing Heterogeneous Sensors and Actuators in Ubiquitous Computing Environments.  In: International Conference on Mobile Computing and Networking, pp. 35--42, Montreal, Quebec, Canada (2007)

[12] Mitra, P., Murthy, C.A., Pal, S.K.: Unsupervised Feature Selection Using Feature Similarity. In: Pattern Analysis and Machine Intelligence, vol. 24, no. 3, pp. 301--312 (2002).

[13] Jiang, W., Yu, D., Ma, Y.: A Tracking Algorithm in RFID Reader Network. In: Japan-China Joint Workshop on Frontier of Computer Science and Technology, pp. 164--171 (2006)

[14] Aydin, E., Oktem, R., Dincer, Z. Akbulut, I.K.: Study of an RFID Based Moving Object Tracking System. In: RFID Eurasia 2007, pp. 1--5, Istanbul (2007)

[15] Soonshin Han, HyungSoo Lim, JangMyung Lee, : 
An Efficient Localization Scheme for a Different-Driving Mobile Robot Based on RFID System. In : RFID Eurasia, 1st Annual, On page(s): 1-5. (2007)

[16] Sangwan, R.S., Qiu, R.G.: Using RFID Tags for Tracking Patients, Charts and Medical Equipment within an Integrated Health Delivery Network. In: 2005 IEEE Networking, Sensing and Control, pp. 1070 --1074 (2005)

[17] Jia, S., Sheng, J., Chugo, D., Takase, K.: Obstacle Recognition for a Mobile Robot Indoor Environments using RFID and Stereo Vision. In: 2007 International Conference on Mechatronics and Automation, pp. 2789--2794 (2007)

[18] Zizka, J., Hroza, J., Pouliquen, B., Ignat, C., Steinberger, R.: The Selection of Electronic Text Documents Supported by Only Positive Examples. In: JADT 2006 (2006)

[19]Lin, C.-H., Liu, J.-C., Ho, C.-H.: Anomaly Detection Using LibSVM Training Tools. In: 2008 International Conference on Information Security and Assurance, pp. 166—171 (2008)

[20] Yufeng Kou, Chang-Tien Lu, Yo-Ping Huang, : Survey of Fraud Detection Techniques. In : Networking, Sensing and Control, IEEE International Conference on, Volume: 2, On page(s): 749- 754 Vol.2. (2004)

[21] TAO GUO, GUI-YANG LI, : NEURAL DATA MINING FOR CREDIT CARD FRAUD DETECTION. In: Proceedings of the Seventh International Conference on Machine Learning and Cybernetics, Kunming, Volume: 7, On page(s): 3630 – 3634. (2008)

[22] Wang, Y., Wong, J., Miner, A.: Anomaly Intrusion Detection Using One Class SVM. In: Fifth Annual IEEE SMC Information Assurance Workshop, pp. 358--364 (2004)

[23] Ghosh, A.K., Schwartzbard, A.: A Study in Using Neural Network for Anomaly and Misuse Detection. In. the 8th conference on USENIX Security Symposium, vol. 8, pp. 12, Washington, D.C. (1999)

[24] Shang-Wei Wanga, Wun-Hwa Chenb, Chorng-Shyong Onga, Li Liuc, Yun-Wen Chuang,：RFID applications in hospitals: a case study on a demonstration RFID project  in a Taiwan hospital. In : Proceeding of 39th International Conference on System Sciences, Hawaii. (2006)

[25] Ho, L., Moh, M., Walker, Z., Hamada, T., Su, C.-F.: A Prototype on RFID and Sensor Networks for Elder Healthcare: Progress Report. In: 2005 ACM SIGCOMM workshop on Experimental Approaches to Wireless Network Design and Analysis, pp. 70-75, Philadelphia, Pennsylvania, USA (2005)

[26] 張維，RFID應用探索，機械工程雜誌249期，pp. 88-94。(2003)

[27] 李和宗，RFID為國內醫療體系來新契機，電子商務導航，pp.6。(2005)

[28] 張映慈，RFID在健康照護的應用，工研院IEK產業資訊。(2004)

[29] Paquet, E., : Exploring anthropometric data through cluster analysis, Published in Digital Human Modeling for Design and Engineering, Seattle, Washington, USA. (2004)

[30] Ethem Alpaydin, : Introduction to Machine Learning, Chapter7, The MIT Press. (2004)