摘要
针对参考独立分量分析收敛速度较慢的问题,提出了两种基于改进的快速收敛牛顿迭代方法的参考独立分量分析方法。该方法首先对观测信号进行白化预处理,避免观测信号矩阵求逆运算,减少了算法的计算量;然后采用修正的具有三阶收敛速度的牛顿迭代方法对参考独立分量分析的代价函数进行优化,推导出快速收敛的参考独立分量分析算法。仿真实验结果表明,改进后的算法是有效的,算法收敛速度相对原算法提高了1.7倍,相对现有算法提高了1倍,而且误差更小。
To overcome the problem that independent component analysis with reference（ ICA-R） has slo-wer convergence speed, two improved independent component analysis with reference algorithms with faster convergence speed are proposed. The new algorithms use the method of pre-whitening to process the ob-served signals to avoid inverse operation of the matrix, and decrease computational time. Secondly, two modified Newton iterative methods with third order convergence are adopted to optimize the cost function of independent component analysis with reference, and deduce the improved independent component analysis with reference. Simulation results prove the efficiency of this new algorithm, and compared with the origi-nal algorithm and the other improved algorithm, the convergence speed of the proposed algorithms raises by 1. 7 times and 1 time respectively with smaller error.
出处
《电讯技术》
北大核心
2014年第1期58-62,共5页
Telecommunication Engineering
基金
吉林省科技发展计划项目(201101110)
吉林市科技发展项目(2013625009)~~
关键词
盲源分离
参考独立分量分析
牛顿迭代
代价函数
收敛速度
blind source separation
independent component analysis with reference
Newton iterative method
cost function
convergence speed
