心拍的镜像高斯建模分析

Mirrored Gauss modeling of ECG beat

心拍分类对于临床心律失常自动化检测非常重要。使用一种新的镜像高斯模型(MGM)算法用于描述QRS复合波段形意,可以自动地、有效地提取QRS复合波段宽度信息,并用于心拍分类。通过使用MIT-BIH心律失常数据库的所有数据集进行测试,正常心拍的总识别率达到93.9%,室性早搏心拍的总识别率达到93.94%。因此,MGM算法可以很好地描述QRS复合波段,并且是一种很有前途的心拍分类算法。

Accurate electrocardiogram （ECG） beat classification is essential for automated detection of arrhythmias. A novel classification algorithm of the ECG beats based on Mirrored Gauss Model （MGM） had been proposed in this paper. The MGM could represent the shape of QRS complex wave. With the MGM, the width of QRS complex wave could be extracted and applied to ECG beat classification easily, effectively and automatically. The experimental results by using all of ECG records in MIT-BIH Arrhythmia Database are that the whole classification accuracy is 93.93% for normal beats and 93.94% for premature ventricular contraction （PVC） beats. Hence, MGM has strong morphological representation ability for QRS complex waves and is a promising algorithm for ECG beat classification.