基于稀疏分解和神经网络的心电信号波形检测及识别  

The Detection and Recognition of Electrocardiogram's Waveform Based on Sparse Decomposition and Neural Network

作　　者：刘金江[1] 王春光[2,3] 孙即祥[3] 

机构地区：[1]南阳师范学院计算机学院,河南南阳473061 [2]空军指挥学院战役训练中心,北京100089 [3]国防科技大学电子科学与工程学院,湖南长沙410073

出　　处：《信号处理》2011年第6期843-850,共8页Journal of Signal Processing

基　　金：河南省科技厅科技攻关项目(NO:102102210483;102102210465)

摘　　要：心电图是现代医学的一个重要诊断依据。用计算机检测识别心电信号波形,能缓解越来越庞大的心电数据给医务人员带来的工作压力,减少因疲劳、疏忽以及主观偏差产生的误差。利用改进的Gabor字典和粒子群优化算法,对心电信号做稀疏分解。稀疏分解得到一个和原信号相比非常稀疏的解向量,与解向量中每个非零值相对应的是从字典中选出的和原信号结构特点最为接近的字典中的一组原子。根据解向量中非零值的大小以及对应原子的波形,确定此原子代表的波的波幅、波宽、波形、位置等信息。然后利用心电信号的先验知识,确定原子代表的波属于那种特征波(P波、QRS波群或T波),进而建立神经网络的训练样本。经过训练,神经网络将能实现对稀疏分解后的心电信号波形的自动检测识别。实验证明,此算法能同时实现几种特征波的检测及识别。Electrocardiogram（ECG） is an important method of diagnosis.Using computer to detect and recognize all kinds of wave of ECG can release doctor＇ s tension coming from more and more ECG data,and decrease error because of tiredness,negligence, and subjective warp.Based on Gabor dictionary modified and the particle swarm optimization,ECG data are decomposed by the theory of sparse decomposition.Comparing of original ECG,its solution vector of sparse decomposition is very sparse（only a few components are nonzero）.Every nonzero value corresponds to an atom selected form dictionary,the atom represents the characters of ECG.Based on magnitude of nonzero value in the solution vector and waveform of atom corresponding,some information,such as amplitude,wave width,waveform,position,can be found out.Then using prior knowledge of ECG,it can be made sure which kinds of waveform represented by atom（P wave,QRS complex or T wave）.And then,it is constructed that training sample of neural network（NN）.Passing through training,NN will complete automatic detection and recognition of ECG.Experiments have proved that this algorithm could complete P wave,QRS complex and T waves＇ automatic detection and recognition of ECG,at the same time.

关 键 词：稀疏分解 神经网络 粒子群优化 心电信号 波形检测识别 

分 类 号：TN911.72[电子电信—通信与信息系统]