基于神经网络和递归模板对准技术的表面肌电信号分解  

Decomposition of Surface Electromyography based on Neural Network and Recursive Template Alignment Technique

作　　者：雷培源[1] 杨基海[1] 赵章琰[1] 魏代祥[1] 

机构地区：[1]中国科技大学电子科学与技术系,合肥230027

出　　处：《生物医学工程研究》2010年第2期84-89,105,共7页Journal Of Biomedical Engineering Research

基　　金：国家自然科学基金资助项目(30870656)

摘　　要：为了提高表面肌电信号(surface electromyography,sEMG)分解的准确率,我们利用空间相邻两通道sEMG信号的信息,采用联合低频小波分解系数作为运动单位动作电位(motor unit action potential,MUAP)活动段的特征,并将自组织特征映射(self-organizing feature map,SOFM)与学习向量量化(learning vectorquantization,LVQ)网络结合起来,完成对MUAP波形的分类。同时为了实现对sEMG信号分解的完整性,采用一种基于递归的模板对准技术分解叠加波形。仿真信号和真实信号的实验表明,本方法具有较高的分解准确率,对于中低收缩力度下sEMG信号的分解十分有效。In order to improve the accuracy of the decomposition of surface electromyography（sEMG）,two-channel sEMG signals adjacent to each other in space were decomposed in this paper.First,the combination of low frequency wavelet coefficients were used as the feature of motor unit action potential（MUAP）.Then,the Self-Organizing Feature Map（SOFM） neural network and the Learning Vector Quantization（LVQ） network are combined together to accomplish the MUAP waveform classification.Simultaneously,to attain completeness of the decomposition,the wavforms superimposed by several MUAP spikes were decomposed adopting the recursive template alignment technique.Experimental results of simulated and real sEMG signals demonstrate that high decomposition accuracies can be achieved using the proposed method,especially for signals recorded at lower to moderate level of contraction force.

关 键 词：表面肌电信号 分解 叠加波形 神经网络 对准技术 

分 类 号：R318[医药卫生—生物医学工程]