請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/60180
標題: | 適用於遠程心電圖監護系統之壓縮感知技術 Compressive Sensing Technologies for ECG Telemonitoring System |
作者: | Yu-Min Lin 林祐民 |
指導教授: | 吳安宇(An-Yeu Wu) |
關鍵字: | 壓縮感知,心電圖壓縮,遠程心電圖監護系統, Compressive Sensing,ECG Compression,ECG Telemonitoring System, |
出版年 : | 2016 |
學位: | 博士 |
摘要: | 隨著高齡化社會以及慢性病增長的趨勢下,居家照護的需求大增。其中如何整合可攜式感測器與無線通訊來實現遠程心電圖監護系統(ECG telemonitoring system),正是實現居家照護的重點方向。目前人體端可攜式無線照護系統面臨到有限的電池容量以及頻寬限制等技術問題,透過心電圖壓縮技術,可以同時解決以上兩個問題,因此開發一個有效的心電圖壓縮系統為本論文主要研究重點。
在遠程心電圖監護系統的訊號處理方面,系統須連續不斷的偵測心電圖訊號以提供即時的病情監控,而眾多的訊號將大量消耗系統的頻寬以及功耗。面對此問題,我們提出利用目前最新的訊號處理技術-壓縮感知(compressive sensing)來解決,壓縮感知技術是2004年由Donoho教授和Candès教授等人在泛函數分析的基礎上,結合了訊號稀疏特性,所提出的突破性訊號取樣理論。壓縮感知可以將高維度的稀疏訊號,透過測量矩陣取得低維度的測量值,因此系統只需要以低維度的訊號做傳遞,等需要時再利用壓縮感知重建演算法將低維度的取樣重建回高維的訊號。利用壓縮感知技術,我們能提供遠程心電圖監護系統更有效率的訊息傳遞方式。 目前壓縮感知技術主要包含三個部分-測量矩陣(measurement matrix)、及最適稀疏基底(proper sparsifying basis)和壓縮感知重建演算法(CS reconstruction algorithm),將壓縮感知技術應用到實際的心電圖壓縮時,其面臨的挑戰包括:1) 現有測量矩陣過於複雜耗費大量記憶體;2) 現有訊號基底未達到最佳稀疏性; 3)現有抗雜訊信號還原演算法運算複雜度過高。 因此本論文提出三項主題已解決上述問題: 首先,本論文提出一個極化循環測量矩陣(polarized circulant measurement matrix),基於循環矩陣所擁有的低記憶體需求特性,提出打亂(scrambling)機制來逼近隨機測量(random measurement matrices)矩陣之效果,並提出其硬體架構設計;比起現有之量測矩陣,所提出之矩陣可以降低65%的功耗。第二部分,本論文提出個人化基底(personalized basis)來基於每個病人的心電圖訊號設計其專屬的基底, 當心電圖訊號在透過基底轉換後越是稀疏時,便可以提高訊號壓縮率,此外搭配個人化基底技術,亦提出電源線干擾去除(interference removal)方法和基底更新(basis refresh)機制;比起現有的多貝西小波基底(Daubechies wavelet basis),所提出的個人化基底可以提高1.8倍的壓縮率。第三部分,本論文提出隨機梯度追蹤演算法(stochastic gradient pursuit)來進行訊號重建,此演算法是一個低複雜度且抗雜訊的重建演算法,我們利用隨機梯度方法會收斂到最小平方誤差解(minimum mean square error solution,MMSE solution)的特性,來抑制雜訊放大,此外我們將提出的還原演算法實現在硬體中,我們提出的還原引擎可以在低成本的情況下達到高吞吐量的特性。 總結本論文所提出之壓縮感知技術,可以有效地解決壓縮感知技術應用到實際的心電圖壓縮時所遇到的問題,並提高壓縮感知技術的應用性。 With the aging of society and rising of chronic diseases, the demand for home care has been increasing substantially. How to integrate the portable sensors and wireless communication to implement the ECG telemonitoring system is the key to home care. Currently, the portable ECG telemonitoring system in the body side is facing the problems of limited battery life and limited bandwidth. ECG compression techniques are employed to address these two issues. As a consequence, our goal is to develop an effective ECG compression system. In the signal processing aspect of ECG Telemonitoring system, the system needs to detect various physiological signals continuously and provides real-time condition monitoring. These numerous signals consume large bandwidth and power in the system. Faced with such dilemma, we propose to exploit the newest signal processing technique, compressive sensing, to resolve these problems. Compressive sensing (CS) is the groundbreaking signal sampling theorem proposed by professor Donoho and Cades based on the functional analysis and sparse signal in 2004. Compressive sensing obtains low dimensional measurements by sampling on high dimensional sparse signals with measurement matrix. Thus, the system only transmits low dimensional signals, while the original high dimensional signals can be recovered by CS reconstruction algorithm. With compressive sensing, we can provide a more efficient way to transmit information in ECG telemonitoring system. Compressive sensing is composed of three core parts, including measurement matrix, determination of proper sparsifying basis, and CS reconstruction algorithm. When CS is applied to the ECG compression, the encountering challenges include: 1) The related measurement matrix is too complex and memory-consuming, 2) The existing sparsifying basis cannot achieve the best sparsity, and 3) The computational complexity of noise-resilient CS reconstruction algorithms is too high. There are three main topics in this work. First, we propose a polarized circulant measurement matrix. Since the hardware cost of circulant matrix is low, we propose a polarization method to scramble the values of traditional circulant matrices to increase randomness and achieve near-optimal performance as Bernoulli random matrices. In addition, architectures are proposed to implement the polarized circulant matrices. Compared with state-of-the-art, the architectures have up to 65% less power consumption. In the second part of this dissertation, we establish a framework on determination of sparsifying basis. We exploit the dictionary learning algorithm to construct personalized sparsifying basis for every patient. As the ECG signals are sparser in sparsifying basis, the compression ratio can be increased. We further propose implicit power-line interference removal scheme and basis refreshing scheme based on the personalized basis. Compared with exiting Daubechies wavelet basis, the proposed personalized basis achieves 1.8 times higher compression ratio. In the third part of this dissertation, we propose stochastic gradient pursuit algorithm for CS reconstruction. It is a low-complexity noise-tolerant CS reconstruction algorithm. The stochastic gradient pursuit algorithm iterates toward a minimum mean square error (MMSE) solution and provides noise suppression. Also, we validate the proposed reconstruction algorithm with chip implementation. Proposed reconstruction engine can achieve high throughput rate under low cost. In summary, the proposed CS technologies can effectively address the encountered problems in CS-based ECG compression. Therefore, the applications of compressive sensing can be extended. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/60180 |
DOI: | 10.6342/NTU201603729 |
全文授權: | 有償授權 |
顯示於系所單位: | 電子工程學研究所 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-105-1.pdf 目前未授權公開取用 | 62.07 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。