利用經驗模態分解法與最小平方法消除心電訊號雜訊

Abstract

心電圖在心血管相關疾病判斷與偵測上扮演相當重要的角色，而一般使用儀器量測到之心電圖通常都會受到雜訊干擾，使得心電圖上各個波形與波段的特徵變得無法辨認與擷取，這對於臨床工作者與醫療人員在病徵上的判斷會造成困擾。雜訊的種類相當多，其中不乏各個頻段的干擾源，主要如：高頻的市電干擾(Power Line Interference)、低頻的基準線漂移(Baseline Drift)與肌肉波(Electromyography, EMG)等皆為心電圖訊號中雜訊的主要來源。因此，心電圖中的雜訊濾除一直是相當重要的課題。 在本論文中，提出了使用經驗模態分解法(Empirical Mode Decomposition, EMD)與最小平方法為基礎的演算法來消除心電圖中雜訊的成分。其中，經驗模態分解法負責訊號分解的部份，經驗模態分解法最大特色就是經由篩選程序能將訊號由高頻至低頻拆解出一序列的振盪訊號，該振盪函數稱為本質模態函數(Intrinsic Mode Functions, IMFs)；訊號重建的部份則是利用最小平方法準則挑選有用的本質模態函數重建無干擾訊號，並利用人為設計訊號來測試此方法的可行性，數值實驗結果驗證了本方法的優點。我們再將此方法應用於自MIT/BIH心律不整資料庫擷取之心律不整實例，並搭配一系列數位濾波器的組合來進行QRS綜合波的檢測，而模擬結果顯示了QRS綜合波的判斷上與MIT/BIH心律不整實例之病徵相吻合，也驗證了經驗模態分解法與最小平方法對於心電圖雜訊濾除的可行性。Electrocardiogram (ECG) has played an important role to diagnose cardiovascular diseases. It often corrupted by interferences introduced by the measurement device. These interferences presented in the signal can lead to the feature of waveforms and frequency bands which can not be recognized and retrieved. These are difficulties for diagnosing symptoms of cardiovascular diseases to clinicians. There are plenty kind of interferences of ECG signals including power line interference, baseline drift and Electromyography, EMG, etc. Thus, the de-noising of ECG is an extremely significant issue. In this paper, a de-noising algorithm based on Empirical Mode Decomposition (EMD) and least square method is proposed to filter the interference of ECG signals. EMD is applied to decompose a signal into a set of oscillatory functions from high frequency to low frequency known as intrinsic mode functions (IMFs) by the sifting process. The interference-free signal is reconstructed by the selected IMFs based on least mean square criterion. Several artificial signals are used as to test the feasibility of the proposed method. Numerical results demonstrate the superiority of the proposed method. This method is also applied to some cases of Arrhythmias from the MIT/BIH Arrhythmias database. Using a set of digital filters’ combination proceed to QRS waves inspections. The simulation results show to conform QRS wave inspections to the symptoms of Arrhythmias and prove the feasibility of the proposed method for processing the ECG signals.