中文摘要：

针对移频轨道电路移频键控信号FSK等的非平稳性,在正交性约束的自然梯度算法的基础上,提出1种利用估计函数的自适应可变步长自然梯度算法,主要原理是根据每次所取分离结果与优化值距离的远近对步长参数取值作在线自适应调整,将每次分离出的瞬时信号的相关性测度作为调整步长参数的依据,同时给出步长参数的上界以使算法稳定。用给出的可变步长算法对FSK信号中的干扰信号进行在线分离。仿真实验结果表明,用可变步长算法分离出的信号精度达到了要求,并且与固定步长算法相比具有更快的收敛速度。

英文摘要：

Based on the orthogonal constraint natural gradient algorithm,an adaptive variable step size natural gradient algorithm using the estimated function was proposed to overcome the non-stationary of the Frequency Shift Keying（FSK）signal.The main principle is that the value of the step size is adaptively regulated online according to the distance between each instantaneous detached signal and the optimal value.The correlation measure of each instantaneous detached signal is taken as the basis for adjusting the step size.At the same time,the upper value of the step size parameter is given to ensure the stability of the algorithm.With the given variable step size algorithm,the interference signal is online removed from the FSK signal.The simulation results show that the precision of the detached signal by the variable step size algorithm meets the requirements.Compared with fixed step size algorithm,the variable step algorithm has higher convergence speed.