基于VMD和KNN的心电信号分类算法

心电信号反映了心脏的活动状态,在诊断心脏疾病和指导心脏手术方面有重要参考价值,为提高心电信号的分类准确率,降低识别难度,提出了一种基于变分模态分解(VMD)和K最近邻(KNN)相结合的心电信号分类算法。使用MIT-BIH心律失常库中的数据,选取5类心电波形作为主要分类对象,从多个记录中截取心电信号作为样本数据,确定最优分解层数,利用VMD分解提取各模态的能量特征作为分类特征,最后运用KNN算法对信号进行分类识别并与支持向量机(SVM)和BP神经网络等分类方法进行了对比。实验结果表明,该方法在少量样本的情况下依然可以实现对心电信号的快速准确分类。

ECG signal classification algorithm based on VMD and KNN

ECG signal reflects the active state of the heart and has important reference value in diagnosing heart diseases and guiding heart surgery,in order to improve the accuracy of ECG classification and to reduce the difficulty of recognition,an ECG signal classification algorithm based on variational mode decomposition( VMD) and K-nearest neighbor( KNN) is proposed. Using the data from the MIT-BIH arrhythmia database,five types of ECG waveforms were selected as the main classification objects,intercepting ECG signals from multiple records as sample data,determining the optimal decomposition layer number and extracting the energy features of each mode as a classification feature by VMD. Finally,the KNN algorithm is used to classify signals and compared with the other methods such as support vector machine( SVM) and BP neural network. The experimental results show that this method can quickly and accurately classify the ECG signal even in the case of smaller number of samples.