A new method for the extraction and classification of single motor unit action potentials from surface EMG signals
A new method for the extraction and classification of single motor unit action potentials from surface EMG signals
复制标题
DOI：
10.1016/j.jneumeth.2004.01.002
复制
发表时间：
2004-07-30
影响因子：
3
通讯作者：
Merletti, R
中科院分区：
文献类型：
--
作者：
Gazzoni, M;Farina, D;Merletti, R
它已被证明，多通道表面肌电图允许评估解剖和生理单个运动单位（MU）的属性。为了获得这些信息，单个MU的动作电位应该从干扰肌电信号中提取。本研究描述了一个自动系统，用于从多通道表面肌电信号中检测和分类MU动作电位。描述了从原始信号中识别和提取动作电位以及将其聚类为它们所属的MU的方法。分割阶段是基于匹配的连续小波变换（CWT），而分类是由一个多通道神经网络，这是一个修改版本的多通道自适应共振理论网络。神经网络可以适应MU动作电位形状的缓慢变化。该方法不需要操作者的任何交互。所提出的技术进行了验证模拟信号，在不同水平的力，由基于结构的表面肌电模型产生。从模拟信号中识别的MU几乎覆盖了整个募集曲线。因此，所提出的算法能够识别代表肌肉的MU样本。从不同的肌肉和条件下记录的实验信号的结果的报告，显示在各种实际情况下的检测到的MU的解剖和生理特性的调查的可能性。该方法的主要限制是，由于MU动作电位叠加，仅在特定情况下才能获得完整的放电模式。(C)2004 Elsevier B.V.保留所有权利。
It has been shown that multi-channel surface EMG allows assessment of anatomical and physiological single motor unit (MU) properties. To get this information, the action potentials of single MUs should be extracted from the interference EMG signals. This study describes an automatic system for the detection and classification of MU action potentials from multi-channel surface EMG signals. The methods for the identification and extraction of action potentials from the raw signals and for their clustering into the MUs to which they belong are described. The segmentation phase is based on the matched Continuous Wavelet Transform (CWT) while the classification is performed by a multi-channel neural network that is a modified version of the multi-channel Adaptive Resonance Theory networks. The neural network can adapt to slow changes in the shape of the MU action potentials. The method does not require any interaction of the operator. The technique proposed was validated on simulated signals, at different levels of force, generated by a structure based surface EMG model. The MUs identified from the simulated signals covered almost the entire recruitment curve. Thus, the proposed algorithm was able to identify a MU sample representative of the muscle. Results on experimental signals recorded from different muscles and conditions are reported, showing the possibility of investigating anatomical and physiological properties of the detected MUs in a variety of practical cases. The main limitation of the approach is that complete firing patterns can be obtained only in specific cases due to MU action potential superpositions. (C) 2004 Elsevier B.V. All rights reserved.