论文部分内容阅读
提出一种基于不完整自然梯度的变步长约束算法,用来处理非平稳环境下的瞬时盲源分离问题.该算法利用系统上的扰动对代价函数进行约束,对算法中的约束因子采用自适应形式,根据分离情况对约束因子进行自适应调整,以加快收敛速度.同时,引入基于代价函数梯度的变步长,使其具有更好的跟踪性能.仿真结果表明,在非平稳环境下,所提出的算法在提高收敛速度的同时可以有效分离源信号而不产生严重的稳态误差.
A variable step size constraint algorithm based on incomplete natural gradient is proposed to solve the transient blind source separation problem in non-stationary environment. The algorithm uses the perturbation on the system to constrain the cost function, Adaptive form, according to the separation of the adaptive constraints to adjust the factor in order to speed up the convergence speed.At the same time, the variable step size based on the cost function gradient is introduced to make it have better tracking performance.The simulation results show that in non-stationary environment, The proposed algorithm can effectively separate the source signal without increasing the steady-state error while improving the convergence speed.