带自适应动量因子的变步长盲源分离方法
发布时间：2019-08-15 20:40
【摘要】：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。
[Abstract]:Based on the natural gradient algorithm, a variable step size blind source separation method with adaptive momentum factor is proposed, which is used for positive definite blind source separation in stationary and non-stationary environments. In this method, the performance index constructor is used to estimate the mixed matrix, and the estimated performance index is obtained according to the estimated mixed matrix, and then the constructor with suitable empirical parameters is replaced by the algorithm, and the step size and momentum factor of the algorithm are adjusted adaptively. Finally, the estimated source signal is obtained. The simulation results show that the mixed matrix can be estimated in both stationary and non-stationary environments, and the mixed signal can be separated effectively and the convergence speed is fast and the steady-state error is small.
【作者单位】： 重庆邮电大学信号与信息处理重庆市重点实验室;
【基金】：国家自然科学基金资助项目(No.61671095,No.61371164,No.61275099) 信号与信息处理重庆市级重点实验室建设基金资助项目(No.CSTC2009CA2003) 重庆市教育委员会科研基金资助项目(No.KJ130524,No.KJ1600427,No.KJ1600429)~~
【分类号】：TN911.7
本文编号：2527208
[Abstract]:Based on the natural gradient algorithm, a variable step size blind source separation method with adaptive momentum factor is proposed, which is used for positive definite blind source separation in stationary and non-stationary environments. In this method, the performance index constructor is used to estimate the mixed matrix, and the estimated performance index is obtained according to the estimated mixed matrix, and then the constructor with suitable empirical parameters is replaced by the algorithm, and the step size and momentum factor of the algorithm are adjusted adaptively. Finally, the estimated source signal is obtained. The simulation results show that the mixed matrix can be estimated in both stationary and non-stationary environments, and the mixed signal can be separated effectively and the convergence speed is fast and the steady-state error is small.
【作者单位】： 重庆邮电大学信号与信息处理重庆市重点实验室;
【基金】：国家自然科学基金资助项目(No.61671095,No.61371164,No.61275099) 信号与信息处理重庆市级重点实验室建设基金资助项目(No.CSTC2009CA2003) 重庆市教育委员会科研基金资助项目(No.KJ130524,No.KJ1600427,No.KJ1600429)~~
【分类号】：TN911.7
【相似文献】
相关期刊论文 前4条
1 马玉梅;武玉厚;;动量因子对BP算法的影响[J];中央民族大学学报(自然科学版);2008年04期
2 杨嘉;杨大力;;MLMS算法中动量因子的分析[J];北京邮电学院学报;1992年04期
3 谭骏;刘辉;;一种动量因子自适应自然梯度的盲分离方法[J];计算机工程与应用;2012年24期
4 ;[J];;年期
，本文编号：2527208
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2527208.html
