摘要
在实际环境中异常值的存在会破坏基于l2范数优化准则的自适应滤波算法,然而在统计学中,基于秩的Wilcoxon方法通常对异常值不敏感,为此,通过以最小化加权Wilcoxon范数代替加权欧式范数作为代价函数,并利用Sign-Sign Wilcoxon比Sign Wilcoxon在存在异常值时有更快的收敛速度这一思想,提出一种基于Wilcoxon范数的变步长符号-符号仿射投影算法,该算法可以缓解基于Wilcoxon范数的符号-符号仿射投影算法固定步长的局限性。仿真结果表明,在非高斯噪声下,该算法与APA,WAPA,Sign WAPA,Sign-Sign WAPA相比具有更优越的性能。
The existence of outliers in actual environment will destroy the adaptive filtering algorithm based on l2 norm optimization criterion.However,in statistics,Wilcoxon method based on rank is usually insensitive to outliers.Therefore,by replacing weighted Euclidean norm with minimized weighted Wilcoxon norm as the cost function,taking the idea that Sign -Sign Wilcoxon has faster convergence speed than Sign Wilcoxon when there are outliers,a new variable step-size Sign-Sign affine projection algorithm based on Wilcoxon norm is proposed,which can alleviate the limitation of fixed step-size of Sign-Sign affine projection algorithm based on Wilcoxon norm.Simulation results show that the algorithm has better performance than APA,WAPA,Sign WAPA and Sign-Sign WAPA under non- Gaussian noise.
作者
李琳琳
郭莹
LI Linlin;GUO Ying(School of Information Science and Engineering,Shenyang University of Technology,Shenyang 110870,China)
出处
《微处理机》
2019年第2期46-51,共6页
Microprocessors