【摘要】 基于十二导联心电图分类任务，提出VGG16与TextCNN融合一维卷积神经网络。VGG16与TextCNN融合模型包括5个卷积层、5个池化层、2个全连接层和1个softmax层。由于十二导联心电图数据规模小，采用传统网络结构与复杂的网络结构得到的结果不佳，为减少模型参数，提高模型计算效率，避免由于模型过于复杂引起的过拟合问题，将十二导联心电信号的十二通道分为12个单通道数据送入VGG16与TextCNN融合模型，基于十二导联心电图进行心率异常二分类研究。在小样本学习的训练过程中，VGG16与TextCNN融台模型平均分类准确率达到了83.64%，能够在样本数量不足的情况下得到较好的准确率，有更为广泛的应用。更多还原

【Abstract】 Based on the duodenal core map classification task,a one-dimensional neural network integrated by the VGG16 and TextCNN is proposed. The VGG16 and TextCNN fusion models include 5 convolution layers,5 pooling layers,2 full connecting layers and 1 SoftMax layer. Since the scale of the duodenal core map data is small,the result of the traditional network structure and the complex network structure is poor. In order to reduce the model parameters,improve the model calculation efficiency,avoid the excessive problems caused by the excessive complexity,the 12-lead electrocardiosignal is divided into 12 single channel data sent into VGG16 and TextCNN fusion model. The dichotomy research of abnormal electrocardigraph based on 12-lead ECG is carried out. During the training of small sample learning,the average classification accuracy rate of VGG16 and TextCNN fusion model reached 83.64%,acquiring good accuracy even when the sample size is small,demonstrating more extensive practical applications.更多还原