论文部分内容阅读
目的研究一种心肌梗死(MI)12导高频心电信号(ECG)全局特征提取方法,以实现阶段性MIECG自动分类识别。方法收集PTB诊断数据库中的健康状态ECG,早期MI ECG,急性期MI ECG,恢复期MI ECG进行研究。提出一种基于联合能量百分比(EP)搜索的二维线性判别法(2D-LDA)对12导高频ECG进行融合特征提取,并进行基于线性分类器的分类。结果各类别分别获得了90.28%～99.24%的分类精度,与常规PCA和LDA法相比,平均分类精度提高了7%～9.7%。结论文中的方法能从12导高频ECG中提取数量较小且分类效果理想的全局心电特征。
Objective To study a global ECG feature extraction method of myocardial infarction (MI) 12-lead high-frequency electrocardiogram (ECG) in order to realize phased MIECG automatic classification and recognition. Methods The health status ECG, early MI ECG, acute MI ECG and convalescent MI ECG were collected from the PTB diagnostic database. A two-dimensional linear discriminant method (2D-LDA) based on joint energy percentage (EP) search was proposed to extract the fused features of 12-lead high-frequency ECG and to classify them based on linear classifier. Results The classification accuracy of each class was 90.28% ~ 99.24%. Compared with the conventional PCA and LDA methods, the average classification accuracy increased by 7% ~ 9.7%. Conclusion The method in this paper can extract a small number of global ECG features with good classification effect from 12-lead high-frequency ECG.