基于深度置信网络的心电信号分类识别

Classification and Recognition of ECG Signals Based on Deep Belief Network

传统的心电信号分类方法通常需要人为提取特征,导致系统的分类性能不稳定。基于此,运用了基于深度置信网络的心电信号分类算法,利用网络的深层次学习能力自动学习信号的特征。提取特征后,选用Softmax分类器对信号进行分类,并用误差反向传播算法微调网络,提高分类性能。选取MIT-BIH数据库中的正常心拍、室性早搏、房性早搏和起搏心拍进行实验,通过实验结果和方法对比,深度置信网络整体的分类精度达到98.8%,表明其在心电信号分类问题中具有良好的分类识别效果。

Traditional ECG classification methods usually need to extract features manually,which leads to the instability of classification performance.Based on this,a ECG signal classification algorithm based on the deep belief network is used.It uses the deep learning ability of the network to learn the characteristics of the signal automatically.After feature extraction,softmax classifier is used to classify the signals,and error back-propagation algorithm is used to fine tune the network to improve the classification performance.The normal beat,ventricular premature beat,atrial premature beat and pacing beat in MIT-BIH database are selected for classification and recognition.Through the experimental results and comparison of methods,the overall classification accuracy of the deep belief network is 98.8%.It shows that deep belief network has a good classification and recognition effect in ECG signal classification.