基於分散式訊源編碼架構的心電圖儀

Abstract

因應高齡化社會的未來趨勢，遠距醫療與健康照護已經成為先進國家重點發展的新興服務產業。心電圖是與心臟搏動相關的電位變化圖，可用來監測病患心臟功能和治療成效。本論文著重於理想傳輸環境下心電圖訊號之壓縮處理，主要是採分散式訊源編碼架構，其壓縮率及合成品質則取決於兩項關鍵元件：訊源相關模型及校驗子生成機制。針對此議題，我們將利用MIT-BIH資料庫的心電圖訊號，並利用向量量化建立匹配的訊源相關模型。除此之外，我們也提出索引層級的軟性輸出解碼演算法，系統模擬證實能有效整合量化索引之事前冗息於解碼演算法。

In recent years, mobile telemedicine has become one of the emerging research topics. Electrocardiogram (ECG) is a pictorial representation of the electrical activity of heart beats and is useful diagnose cardiac disease. The purpose of this work is to develop distributed source coding (DSC) techniques for ECG signal compression under an ideal transmission environment. We focus on the DSC of two correlated ECG sources, with their correlation modeled by a virtual binary symmetric channel (BSC). Based on the concept of code binning, the DSC encoder compresses one source sequence into a syndrome sequence. The DSC decoder uses the other source sequence as side information together with the received syndrome sequence to identify the original source sequence. Also proposed is a modified BCJR algorithm which integrates the symbol level a priori information into the soft-output decoding algorithm.