基于改进深度残差网络的心电信号分类算法

ECG signal classification algorithm based on improved depth residual network

针对不同类型心电(electrocardiogram,ECG)信号分类的不足,提出了一种基于改进深度残差网络(Resnet)的分类方法。首先对心电数据做可视化处理,使用格拉姆角场(Gramian angular fields,GAF)将一维的ECG信号转换为二维图像,然后对Resnet-50网络模型进行改进,在网络中添加多级shortcut支路,并优化了残差块;为了进一步提高模型的表达能力,将Relu激活函数替换为SELU激活函数;最后将图像输入到改进的残差网络中进行分类,并在医院对患者的心电信号进行了实际测试。实验结果表明:该算法对7类心电信号的平均识别率达到了98.3%,相对于原始的残差网络,准确率提升了2.9%;算法诊断结果与医生诊断结果一致,从而验证了算法的有效性和实用性。

Aiming at the deficiency of classification of different types of electrocardiogram(ECG)signal,a classification method based on the improved depth residual network(Resnet)is proposed.Firstly,the ECG data is visualized,and the one-dimensional ECG signal is transformed into a two-dimensional image using the Gramian angular fields(GAF).Then,the Resnet-50 network model is improved by adding multi-level shortcut branches to the network and optimizing the residual blocks.In order to further improve the expressiveness of the model,the Relu activation function is replaced by the SELU activation function,and finally the image is entered into the improved Resnet network for classification.The ECG signals of the patients were actually tested in the hospital.The results show that the average recognition rate of the algorithm for 7 kinds of ECG signal reached 98.3%,compared with the original residual network,the accuracy rate is improved by 2.9%;the diagnostic results of the algorithm are consistent with those of doctors,which verifies the effectiveness and practicability of the algorithm.