本論文主要研究目的是建立一套快速篩選技術,達到即時判讀心音的系統,並針對正常的心音、心雜音與射血音、迴血音訊號進行分類與辨識。 心雜音的診斷過程,最簡易的方法之一為透過心臟聽診,直接經由臨床醫師依聽診經驗進行診斷,其他方法例如使用心臟超音波,亦可以達到進一步的診斷目的,但是這類儀器的體積龐大,需要許多的配合才可使用,並不容易在第一時間使用來辨識心雜音。因此,本研究針對臨床主要之幾種心雜音病症,收錄心房中隔缺損、法洛氏四重症、開放性動脈導管、主動脈瓣狹窄、肺動脈狹窄、心室中隔缺損的病患與正常人無心臟雜音的心音訊號,進行初步的分析研究。本論文之研究方法是在擷取受測者心音訊號後,利用二次雲線小波(Quadratic Spline Wavelet)分析,最後再對心音訊號做定量化分析,找出心音訊號的差異性。由分析結果得知,本論文可利用心音訊號的靜音與雜音的能量,區別出正常心音與具有心雜音的訊號,更進一步的將具有心雜音的訊號分類為射血性心雜音與迴血性心雜音。另外,在正常人的心音與具有雜音的心音訊號上,成功的不依賴ECG參考訊號,將第一心音的位置明確定位,並依此切割心音訊號,本論文之方法針對第一心音的定位正確率可達96.9%。
The major purpose of the thesis is to build a system of fast screening technology to diagnose real-time heart sound and to aim at classifying and recognizing normal heart sound, murmur, ejection murmur and regurgitant murmur. In a process of diagnosing murmur, one of the simplest methods is to diagnose via cardiac auscultation by physicians according to their experiences. Other methods like using cardiac ultrasonography can also achieve the same diagnosis, but the size of this kind of instrument is large and numerous matches are needed. On the top of that, it could not recognize murmur in the first time. Therefore, the thesis focuses on mainly several murmurs in clinic such as ASD, TOF, PDA, AS, PS, VSD and record heart sound signals of these murmurs and normal heart sounds to process preliminary analysis. The study method of the thesis is to acquire objects’ heart sound signals by using quadratic spline wavelet analysis. Finally, find out the difference of heart sound signals by quantitative analysis. From the analyzed result, the thesis can use the energy of silence and murmur of heart sound signals to distinguish between normal heart sounds and murmurs and can further classify murmur’s signals into ejection murmur and regurgitant murmur. Besides, the method of this thesis can successfully orientate first heart sound in signals of normal heart sound and murmur without depending on ECG signals and to segment heart sound signals by the result of orientation. The accuracy of orientate first heart sound in the thesis is 96.9%.