Research about Motor Unit Visualization with Surface EMG Signals

A)运动单元分解技术的改进我们提出了一种新的基于超完备基的运动单元分解技术，引入了统计模型。经验证，该方法应用于实际测量的表面肌电信号，具有与盲解卷积相当的分解性能。B)利用统计模型检验识别技术的有效性。提出了一种基于KDA和Boost机制的识别系统，并验证了该系统的识别性能与支持向量机等已有技术的识别性能相当。一般而言，为了最大化泛化性能，参数调整过程，如交叉验证，其计算…更多的成本是非常昂贵的，是需要的。为了解决这个问题，我们提出了新的参数选择指标值，利用这个指标，可以在不增加昂贵的计算成本的情况下选择合适的参数。c)每个分解的运动单元的三维位置估计利用三维有限元方法从皮肤表面的电位分布估计单个运动单元的去极化的三维位置。结果证实，可以估计单个运动单位多极化的三维位置和动力学。此外，还可以估计神经支配区域的大小和电流去极化强度的时间动态。此外，这项研究是在神户大学助理教授、首席研究员小谷博士的2003财年开始的。然而，他在2004年5月去世，所以头像突然被换了。我写下Kotani博士的贡献涵盖了上述结果的全部内容。较少

This research was executed for the motor unit decomposition and making the activity visible from the surface electromyogram as follows.A)Improvement of motor unit decomposition techniqueWe proposed a new motor unit decomposition technique with overcomplete bases to introduce the statistical model. It was confirmed that this technique was applied to the real measurement surface EMG signals, and it had the decomposition performance equal with blind deconvolution. Moreover, it was confirmed that each motor unit activity might be separable even if the number of observation channels were less than the number of active motor units.B)Examination of effectiveness of recognition technique using statistical modelWe proposed the recognition system using both KDA and boosting mechanism and confirmed that it's recognition performance was just like an existing technique such as SVM. In general, to maximize generalization performance, parameter tuning process such as cross validation, whose computati … More on cost is very expensive, is needed. To solve this problem, we proposed new index value for parameter selection, and with this index, appropriate parameters can be selected without so expensive computation cost.C)The three dimensional position estimation of each decomposed motor unitThe 3D position of the depolarization of individual motor unit was estimated by using the 3D finite element method from the potential distribution on the skin surface which was estimated with motor unit decomposition technique from surface EMG signals. As a result, it was confirmed that 3D position and dynamics of dopolarization of individual motor unit might be estimated. Moreover, the size of innervation zone and temporal dynamics of current intensity of depolarization could be estimated.In addition, this study had been started in 2003 fiscal year under Dr.Kotani head researcher, Assistant Professor of Kobe University. However he died in May, 2004, so the head was changed suddenly. I write down that contribution of Dr.Kotani covers the whole of the above results. Less