一种心音小波神经网络识别系统  

A recognition system for a heart sound wavelet neural network

作　　者：成谢锋[1,2] 傅女婷 陈胤[1] 张学军[1,2] 黄丽亚[1] 

机构地区：[1]南京邮电大学电子科学与工程学院,南京210003 [2]南京邮电大学射频集成与微组装技术国家地方联合工程实验室,南京210003

出　　处：《振动与冲击》2017年第3期1-6,共6页Journal of Vibration and Shock

基　　金：国家自然科学基金(61271334;61373065)

摘　　要：设计一种心音小波神经网络识别系统,将心音特征抽取、有针对性的神经网络层次化架构和分类识别融合一体,以解决复杂条件下的心音分类识别问题。提出基于心音小波神经网络的识别模型,讨论如何构造心音小波和心音小波神经网络的方法,重点讨论在网络结构的隐含层中引入心音小波作为激活函数的算法,从而获得一种把心音的针对性学习和心音识别技术高度融合的心音小波神经网络识别系统。通过选取正常心音信号与早搏心音信号作为实验对象,验证了心音小波神经网络识别系统的有效性和实用性,并且通过与morlet和Mexican-hat小波神经网络识别系统相比较,证明心音小波神经网络识别系统在收敛性、算法速度上呈现明显的优越性。Here, a recognition system for a heart sound wavelet neural network was designed, with it heart sound features were extracted, the neural network structure was layered, identified and classified to solve the problem of heart sound classification recognition under complicated conditions. Firstly, a recognition model based on heart sound wavelet neural network was proposed. Then the method to construct heart sound wavelet and heart sound wavelet neural network was discussed. The algorithm for introducing heart sound wavelet as the activation function of neural network hidden layer was discussed specially to obtain a recognition system of heart sound wavelet neural network highly fusing heart sound targeted learning and heart sound recognition technique. Finally, normal heart sound signals and beats heart sound signals were selected as the test objects and to verify the effectiveness and practicality of a heart sound wavelet neural network recognition system. Compared with Morlet wavelet neural network recognition system and Mexican hat wavelet neural network recognition system, it was shown that the heart sound wavelet neural network recognition system is superior in convergence and alrorithm speed.

关 键 词：心音 识别 心音小波神经网络 

分 类 号：TP11[自动化与计算机技术—控制理论与控制工程]