基于分值函数估计的自然梯度盲分离算法  

Blind Source Separation Algorithm of Natural Gradient Based on Estimation of Score Function

作　　者：张延良[1,2] 楼顺天[1,3] 张伟涛[1] 常华[1] 

机构地区：[1]西安电子科技大学电子工程学院,西安710071 [2]河南理工大学计算机科学与技术学院,焦作454001 [3]超高速电路设计与电磁兼容教育部重点实验室,西安710071

出　　处：《数据采集与处理》2011年第2期167-171,共5页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金(61074102)资助项目

摘　　要：分值函数估计的准确性是影响自然梯度盲分离算法收敛速度和稳定性的一个重要因素。提出了一种通过函数逼近直接估计分值函数的方法。用一组正交多项式的线性组合来逼近分值函数,其性能用均方误差来衡量。线性组合的系数向量可运用分值函数的性质,通过最小化均方误差自适应学习得到。将估计到的分值函数代入自然梯度迭代公式中就得到新算法。实验仿真表明,与传统的自然梯度盲分离算法相比,新算法的收敛速度与稳定性均有显著提高。The accuracy of estimated score function is an important factor that influences the convergent speed and stability of the natural-gradient-based blind source separation（BSS） algorithm.A new estimating score function approach based on the method of function approximation is proposed.In this approach,score function is approximated by the linear combination of a set of orthogonal polynomials.The accuracy of approximation is measured by mean square error（MSE）.Using the property of score function,coefficients of the linear combination can be obtained by adaptively minimizing MSE.The new BSS algorithm is developed by substituting the estimated score function into natural gradient iterative formula.Experimental results show that,compared with the traditional one,the stability and convergent speed of the new algorithm is improved.

关 键 词：盲信源分离 独立成分分析 分值函数 函数逼近 自然梯度 

分 类 号：TN911.7[电子电信—通信与信息系统]