【摘要】 基于混沌动力学系统相空间的延迟坐标重构及人工神经网络的非线性特性。研究了采用基于自适应投影学习算法的径向基函数网络对实测的EEG信号进行预测。通过对径向基函数引入一宽度调节系数α ,使网络的预测性能有较大提高。理论分析和研究结果表明 :α的取值由EEG信号的关联维数D2 决定 ,α在最佳区间内取值能够更有效地对EEG信号进行预测。更多还原

【Abstract】 Based on the delay time phase reconstruction of the chaotic dynamical system and the nonlinear characterization of the neural networks, the prediction of factual EEG signals were investigated by using the radial basis function neural networks with the adaptive projective learning algorithm. By introducing a parameter α to adjust the width of the original radial basis function, the predictive characteristics of the network were improved greatly. Theoretical analysis and experimental results show that the value of the parameter α is determined by the correlation dimension (D 2) of the EEG signal. The EEG signals can be predicted more efficiently if the value of α is selected in the optimal region.更多还原