摘要
目的提出了一种基于模板匹配和特征识别相结合的波形分类算法 ,用于心室期前收缩波形的分类。方法通过自适应的可变区域模板匹配技术从形态学上综合地比较了正常和异常的波形 ,用匹配结果指导后续的特征分类。特征分类过程仅使用 4个参量 ,基于最大概率分布的模糊识别解决了参量分布的模糊性问题。结果经MIT/BIH标准心电数据库测试 ,实验结果达到了 99.5 1 %的特异度和96.84%的灵敏度。结论该算法在分类精度上较已有的算法有明显的提高。
Objective A beat classification algorithm based on the combination of template matching and characteristic recognition is presented and applied on the Premature Ventricular Contraction beat classification. Method The adaptive Variable Area Template matching technique was used to synthetically introduce morphological information of normal and abnormal waveforms. The matching result directed the subsequent characteristic classification. Only four features were extracted for classification, and the fuzzy recognition according to maximum probability represented the fuzzy distribution of features. Result The performance of the algorithm was evaluated on the MIT BIH arrhythmia database and the result showed that the specificity and the sensitivity for PVC reached 99.51% and 96.84% respectively. Conclusion The precision of the proposed algorithm was evidently higher than those of the existing algorithms.
出处
《航天医学与医学工程》
CAS
CSCD
北大核心
2002年第2期98-102,共5页
Space Medicine & Medical Engineering
关键词
心室期前收缩
模板匹配
特征识别
可变区域模板
特征分类
venricular premature contraction
template matching
characteristic recognition
variable area template
characteristic classification