带自适应动量因子的变步长盲源分离方法

Variable-step blind source separation method with adaptive momentum factor

基于自然梯度算法提出一种带自适应动量因子的变步长盲源分离方法,在平稳和非平稳环境下进行正定盲源分离处理。该方法利用性能指标构造函数来估计混合矩阵,依据估计混合矩阵得出估计性能指标再反馈更新构造函数;然后将选取合适经验参数的构造函数代入算法,同时自适应调整算法步长和动量因子;最终得到估计源信号。仿真表明该方法在平稳和非平稳环境下都可以估计出混合矩阵,能有效分离混合信号且收敛速度快稳态误差小。

A variable-step blind source separation algorithm based on the natural gradient with adaptive momentum factor was proposed,which could cope with the determined blind source separation in the environment of stationary and non-stationary.Function estimation mixed matrix was constructed by performance index.The estimated performance index was obtained by the estimated mixed matrix,and the constructor was updated by the estimated performance index.Then,the constructor was plugged with appropriate experienced parameter into the proposed algorithm and step and momentum factor was adaptively adjusted.Finally,the estimation source signals could be obtained.Simulations show that the proposed algorithm is effective to estimate the mixed matrix in the stationary and non-stationary environments,and the proposed algorithm has faster convergence speed and lower steady error as well as separates source signals effectively.