论文部分内容阅读
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。
ECG is an important diagnostic basis for cardiovascular disease. To explore new methods to deal with ECG signals has important theoretical and practical value for medical diagnosis and treatment. An adaptive neuro-fuzzy inference system (ANFIS) with five-layer structure network including input node layer, regular node layer, average node layer, conclusion node layer and output node layer is presented. The hybrid adaptive learning algorithm based on Sugeno fuzzy theory, least square method and gradient descent method is used to train the parameters of neural network in ANFIS to improve the convergence performance of ANFIS system. To verify the effectiveness of ANFIS in the detection of ECG signals, the network training of ANFIS was carried out through the first CECG signal and the last ECG signal (MECG) in the measured data of the original ECG signal. The experimental results show that the adaptive neural network fuzzy inference system is very effective in the analysis and prediction of ECG signal.