摘要
目的研究适于心搏分类的突出特征矢量。方法首先采用连续小波变换对QRS复合波进行定位,然后采用不同的特征提取技术,提取一组特征矢量,送入线性判别式分类器进行训练,并对基于MIT/BIH数据库中的4类心搏进行分类,评价其分类性能。结果用10维特征矢量对4类心搏进行分类,准确度可达97.83%。结论综合利用不同特征提取技术可以显著提高心搏分类的准确度。
Objective To investigate pertinent features fitting for beat classification.Methods The QRS complex was firstly detected by using continouse wavelet transform（CWT）,then the feature sets were detected by different feature extraction technique.Using linear discriminants,4 types of ECG beats based on MIT/BIH arrhythmia database were classified and the classification performance of the feature sets was evaluated.Results A feature sets of 10 dimensions with the accuracy of 97.83 % was achieved.Conclusion The classification accuracy is significantly increased in detection with combining different feature extraction technique.
出处
《生物医学工程与临床》
CAS
2007年第5期344-347,共4页
Biomedical Engineering and Clinical Medicine
基金
国家自然科学基金资助项目(60661002)
