基于FastICA和通道间相关的表面肌电信号分解研究

目的为了对神经肌肉疾病进行相关的研究和临床上诊断治疗,探索新的和有效的表面肌电(surface EMG,sEMG)信号分解方法。方法首先用FastICA求解混矩阵,然后对测量信号矩阵进行变换,再用通道间相关性分解s EMG信号。结果经过仿真和真实信号进行测试,分解信噪比为0 d B的第一组信号时,以平均95.6%的准确率分解出20个运动单元(motor unit,MU);分解信噪比为20 d B,且参与发放的MU更多,发放频率更高的第二组信号时,以平均98.4%的准确率分解出29个MU;分解真实信号时,得到的平均MU个数为14.2,并用"二源法"进行评测,两组中分解出相同MU的比例为80%,且相同MU发放时刻的平均重合率为95.1%。结论这种结合Fast ICA和通道间相关的方法能以较高的准确率实现s EMG信号的有效分解。

Research on Decomposition of Surface EMG Signals Based on FastICA and Channel Correlation

Objective A new and effective method for decomposing surface electromyography (sEMG) signals was explored for the relevant research,diagnosis and treatment of clinical neuromuscular diseases.Methods The FastICA method was employed to obtain the de-mixed matrix which was then used to transform the matrix of measurement.At last,the sEMG signals were decomposed by utilizing the channel correlations.Results Two groups of simulation signals and one group of real sEMG signals were tested and the results showed that for the first group of simulation signals with 0dB SNR,an average of 20 motor units were extracted with an average accuracy of 95.6％;While for the second group of simulation signals with 20 dB SNR,an average of 29 motor units were extracted with an average accuracy of 98.4％.As to the real sEMG signals,an average of 14.2motor units were extracted.A "two-source" test was further conducted to evaluate the performance of the proposed method,it showed that the proportion of the same MU extracted by both groups was 80％,and the average coincidence rate of the same MU was 95.1％.Conclusion The combination of FastICA and channel correlation method can effectively decompose the surface EMG signals with high accuracy.