改良式K-means演算法於監測照護系統之圖樣識別

A Modified K-means Algorithm for Pattern Identification on Monitoring Care Systems

張雅筑(Ya-Chu Chang)

指導教授 ： 練光祐

國立臺北科技大學/電資學院/電機工程系研究所/碩士(2011年)

https://doi.org/10.6841/NTUT.2011.00344

摘要

因應高齡化社會的趨勢，以輔助科技來改善高齡者之照護逐漸受到重視。本研究針對睡眠照護議題，設計一套自動監測系統，監測高齡者於臥床、離床、翻身等姿勢變化，藉此得知是否在床上睡得安穩。本文提出一套改良式K-means演算法，用於監測系統的圖樣識別上。基於傳統K-means演算法的分群方式，將必須給定之群集數目改為透過合理性指標判斷出合適群集數目，以及改變初始群集中心點產生的方法，降低原本因隨機產生而造成分群結果的不確定性。研究方式以硬體建構一套按壓式感測面板並做為姿態圖樣的採集訊號輸入端，訊號接收端使用National Instruments的DAQ裝置將訊號傳送至電腦，對應的監控系統建構於LabVIEW，設計一套能即時觀測姿態圖樣與分析圖樣特徵點的人機介面，將感測面板與人機介面組成監測模組，用以實現監測照護系統。最後，進行模擬監測身體姿勢之實驗以驗證本文提出之改良式K-means演算法對分群結果的合理性，以及將分群結果定義為即時判讀監測姿態的依據。

關鍵字

K-means分群演算法 ； 群集數目 ； 初始群集中心點 ； 監測系統 ； LabVIEW

並列摘要

In response to the tendency of aging society, this thesis focuses on assistive technology to improve the monitoring care systems of the elderly. We develop an automatic technology to monitor sleeping posture. The variation of the sleeping posture of human being is closely related to their physiological status. To this end, we propose a modified K-means algorithm for pattern identification on monitoring sleeping posture. Based on the general K-means algorithm, we improve the clustering results through the proposed algorithm. We used the algorithm to find an appropriate number of clusters and produce the suitable initial center of each cluster. We construct a pressure button sensor panel to mimic a bed with dedicated pressure sensors. The monitoring platform is built on National Instruments LabVIEW and its function contains signal acquisition, analysis and processing. Finally, we carry out many monitoring experiments and verify our algorithm with satisfactory results.

並列關鍵字

K-means Clustering Algorithm ； Number of Clusters ； Initial Cluster Center ； Monitoring System ； LabVIEW

參考文獻

[30] 集研科技股份有限公司，http://www.systemaccess.com.tw/。

[4] J.A.M. Graafmans and H. Bouma, ”Gerontechnology, fitting task and environment to the elderly,” Human Factors and Ergonomics Society, 1993, pp.182-186.

[6] I. Davidson, ”Understanding K-means non-hierarchical clustering,” SUNY Albany Technical Report, 2002, pp.2-25.

[7] A. Ahmadyfard and H. Modares, “Combining PSO and K-means to enhance data clustering,” International Symposium on Telecommunications, 2008, pp.688-691.

[9] J. Xie, Y. Zhang and W. Jiang, “A K-means clustering algorithm with meliorated initial centers and its application to partition of diet structures,” International Symposium on Intelligent Information Technology Application Workshops, 2008, pp.98-102.