摘要
目的提高心电信号的分类准确率,降低算法复杂度。方法首先以MIT-BIH心电数据作为学习模板,然后在心电信号的频域和时域上提取其离散余弦变换(discrete cosine transform,DCT)、RR间期和QRS复合波的三种特征值进行分析,最后采用最小欧式距离分类器判断待测心电信号的类型。结果该分类模型通过MIT-BIH和AHA国际标准心电数据库的验证,分别得到96.6%和94.1%的分类准确率。结论本文的心电分类模型区别于其他分类算法的一个最大特点就是算法复杂度低,这是异常心律能够被实时检测和预警的关键,而且建立的心电分类模型已经能够在普通的手机平台上实现。
Objective To improve the classification accuracy of ECG and reduce the complexity of the algorithm. Methods This paper uses MIT-BIH ECG database as learning templates ,then extracts its eigenvalues of DCT, R-R interval and QRS from frequency and time domain of ECG signals to analyze. Finally,the type of ECG signal is classified based on the minimum Euclidean distance classifier. Results The classification model is tested and verified by international standard MIT-BIH and AHA ECG database,with the classification accuracy of 96. 6% and 94. 1%, respectively. Conclusions Lower complexity in ECG classification model than other algorithm is the greatest feature, which is the key of detecting real-time abnormal heart rhythms.' And ECG classification model has been realized on a common mobile platform.
出处
《北京生物医学工程》
2016年第3期259-266,共8页
Beijing Biomedical Engineering
关键词
离散余弦变换
特征分析
最小欧式距离
discrete cosine transform
feature analysis
minimum euclidean distance