基于自适应神经网络模糊推理系统的心电信号检测  

Electrocardiograph Signal Identification Based on Adaptive Neuro-Fuzzy Inference System

作　　者：盛维涛[1] 张文君[1] 袁宇鹏[2] 苏航[2] 

机构地区：[1]四川工程职业技术学院,四川德阳618000 [2]重庆大学自动化学院,重庆400044

出　　处：《重庆师范大学学报（自然科学版）》2015年第6期140-144,共5页Journal of Chongqing Normal University:Natural Science

摘　　要：心电信号是心血管疾病的重要诊断依据,探索新方法来处理心电信号对于医学诊疗具有重要的理论意义与实用价值。阐述了一种包含输入节点层、规则节点层、平均节点层、结论节点层和输出节点层的五层结构网络的自适应神经网络模糊推理系统(Adaptive neuro-fuzzy inference system,ANFIS),并提出了基于Sugeno模糊理论、最小二乘法和梯度下降法的混合自适应学习算法来训练ANFIS中的神经网络的参数,来提高ANFIS系统的收敛性能。为验证ANFIS系统在心电信号检测中的有效性,通过原始心电信号的实测数据中的第一路腹壁混合信号(CECG)和最后一路母体心电信号(MECG)进行了ANFIS的网络训练,基于训练结果对于腹壁混合信号进行了实验预测分析,实验结果表明自适应神经网络模糊推理系统在心电信号的分析与预测中十分有效。Electrocardiograph（ECG）signal is significant for the judgment of cardiovascular disease.It is great importance in studying the novel approach to deal with the electrocardiograph signals.An adaptive neuro-fuzzy inference system（ANFIS）is presented in this paper to process the electrocardiograph signal,which consists of five network layers：input node layer,rule layer,average node layer,conclusion node layer and output node layer.An intelligent hybrid learning algorithm is proposed to identify the parameters of the ANFIS,which is mixed with the Sugeno fuzzy theory,least squares and gradient descent method.The effectiveness and convergence of the ANFIS can be improved by the hybrid algorithm.Experiment is carried out to test the estimation performance of ANFIS.Two channels of electrocardiograph signal are utilized as the system inputs of the ANFIS,which are the abdominal composite signal（CECG）and maternal electrocardiograph signal（MECG）.ANFIS is applied to predict the CECG by the primary training.Experimental results demonstrate that the ANFIS has good performance in predicting the electrocardiograph signal.

关 键 词：自适应神经网络模糊推理系统 心电信号 Sugeno模糊理论 最小二乘法 梯度下降法 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]