基於二導程心電圖的壓縮域身份辨識系統

Abstract

在無線遠程醫療應用中，高效率的傳輸和儲存需求有賴於心電圖的資料壓縮處理，身份辨識必須執行解壓縮後，才能相容於現有運作在原始域的生物辨識技術。為了實現更迅速的病人照護，本論文提出了一種運作於壓縮域的心電圖生物特徵辨識方法，在JPEG2000架構上將身份辨識簡化成一項內涵式影像資訊檢索問題。有別於前人研究大多選擇單一肢導程作為訊號來源，本論文加入心前導程訊號，有效改善單一導程心電圖對心臟疾病患者身份辨識效果不佳之瓶頸。首先，將二導程一維心電圖訊號經過預處理程序轉變成二維心電向量圖及心電圖像，再進行JPEG2000影像編碼標準的壓縮處理。透過此機制將心電圖的形態特徵轉換為灰階影像，再利用其形狀和紋理特徵作為個人身份辨識的依據。由實驗結果顯示，本論文提出的壓縮域心電圖生物特徵辨識系統，在兼有健康受測者與心律不整患者的141人心電圖資料庫測試具有96.70%的辨識精準度。

In wireless telecardiology application, compressed ECG packets are often required for efficient transmission and storage purposes. Compressed ECG data must be decompressed first before applying existing biometric techniques that work on the original signal. In this work, we propose a new method which applies two-lead ECG signals to perform person identification in compressed domain. Incorporation of compressed domain ECG biometric enables faster person identification as it bypasses the full decompression. Compared to previous studies that focus on one limb lead, this study adds precordial lead as another source for person identification. Experiment results show that the combined use of two-lead features can improve the identification accuracy for the arrhythmia patients. First, we apply a preprocessor which converts two-lead 1-D ECG signals to 2-D image matrices and compress them by JPEG2000 image coding standard. Feature relating to ECG morphology were then extracted for indexing person identity by shape and texture content in a known enrollment database. Experiments on public ECG databases demonstrate the validity of the proposed method for biometric identification with an accuracy of 96.70% on both healthy and diseased subjects.