基于RBF神经网络的肌电信号建模仿真研究

Radical Base Function Neural Networks-Based EMG Modeling

建立肌电信号模型 ,尤其是肌肉单纤维动作电位模型 ,并进行计算机仿真研究 ,是研究从复杂信号中提取特征参数新算法性能的基础。利用反向传播算法网络和径向基函数的强逼近能力 ,建立模拟合成单纤维动作电位的精确模型 ,以达到快速、准确地计算机合成单纤维动作电位波形。通过实验比较 ,径向基函数网络较反向传播算法网络更适合于肌电信号这类非线性系统建模。

It is the basis in studying the performance of the new methods for picking-up characteristic parameters from complicated signal that set up EMG signal, especially MFAP model. In order to simulated synthesize MFAP waveform fast and truly this paper set up an exact MFAP simulated model based on the strong approaching capability of RBF networks. Through comparing with BP networks performance, we found that RBF networks achieve better result in and are suitable for modeling a non-linear system such as EMG.