摘 要 本文旨在自行建立一套心電訊號擷取系統，利用此系統獲取數位化的心電訊號，並於電腦上做後續數位處理，如數位濾波、心電訊號平滑化和QRS波偵測等。此系統主要量取誘導II及誘導III，並應用愛因託芬氏三角形計算出其它心臟額平面誘導，分別為誘導I、右手加壓誘導(aVR)、左手加壓誘導(aVL)及左腳加壓誘導(aVF)等。 於訊號擷取系統中，由於類比濾波器對於雜訊無法完全抑制，因此需要後續利用數位濾波來消除心電訊號中的雜訊，研究中利用四種不同數位濾波進行雜訊消除，並對其優缺點進行比較。為了達到病徵診斷的目的，心電訊號峰值的偵測極為重要，研究中對P波、QRS波及T波分別進行偵測，以便後續專家系統發展。ㄧ般而言，計算心跳速率均是利用R-R波的間隔時間來推算出心跳的變化，本文提出應用時域脈動與頻譜關聯性於心電訊號，從頻譜圖找出心跳頻率並反算其心跳率。 為了要獲得平滑的心電訊號，一般會使用平均法將較低能量的高頻雜訊濾除，但是使用此方法將造成心電訊號中R波衰減，導致心電訊號有不正確的訊息，即使是安捷倫V24生理監視器也都有如此情況。因此本研究提出一個新的改進技術，實際應用於心電訊號的平滑化。這個新的技術主要利用Gabor濾波器將心電訊號平滑化，而且平滑後的心電圖能夠保留原有訊息，相較於傳統平滑化的缺點有大幅的改善。最後於研究中，將使用頻譜濾波器及Gabor濾波器來獲得較完整及較平滑的心電訊號，並對其峰值作偵測以獲得心電圖的相關訊息，如心跳率(亦於頻譜中找出)、PR波間距等，作為以本系統進行心電訊號處理之具體建議。 Abstract The main purpose of this study is to design an ECG measuring system for investigating the electrical activity of heart. At the measuring system, two frontal plane limb leads, lead II and III, are measured. The other four frontal plane leads can then be calculated by using the Einthoven triangle which is lead I, aVR, aVL, and aVF, respectively. The acquired analogue ECG signals are still not acceptable, further digital filter is needed to obtain improved illustration for clinic diagnosis. In the study, we provide four different types of filter to remove noise and compare their performance. To detect ECG peaks, we compute some information of ECG signals for diagnosis. In general, the heart rate computation is to employ R-R interval. But another method detecting heart rate in the spectrum is proposed in the study. One of the ECG signal processing targets is to smooth the data to reduce high frequency noise and to improve SNR in signals. The conventional technique can cause R-wave peaks reduction, such as using Agilent component monitoring system V24. Hence, the Gabor filtering technique is firstly proposed and implemented to cope with this drawback without missing any tiny information. Finally, we apply the spectrum and Gabor filters to obtain smoothed ECG signals for clinic or diagnostic purpose. Some information of an ECG such as heart rate and PR wave interval is computed by peak detection through proposed spectral analysis. They conclude the ECG signal processing in the study.
