广义线性结构广泛用于解决共平台系统中同相/正交相位(in-phase/quadrature-phase,IQ)不平衡引起的自干扰镜像信号残留问题。由于在共轭项维度上的拓展,输入信号间的相关性使得用于实现权值向量自适应更新的广义线性复数最小均方(widely...广义线性结构广泛用于解决共平台系统中同相/正交相位(in-phase/quadrature-phase,IQ)不平衡引起的自干扰镜像信号残留问题。由于在共轭项维度上的拓展,输入信号间的相关性使得用于实现权值向量自适应更新的广义线性复数最小均方(widely linear complex least mean square,WLCLMS)算法出现明显的收敛速度下降。针对这一问题,本文提出了一种二维正交化方法,通过特征值分解实现输入信号在样本延时以及共轭项两个维度上的去相关。同时建立了包括格形预测器、基于梯度的白化器,以及数字对消器三级结构的自适应二维正交化WLCLMS模型,实现了时变环境中自干扰信号的实时跟踪。仿真结果表明,该方法在有效提升收敛速度的同时,有着稳定的稳态误差性能,并且针对系统采样率、滤波器阶数、IQ不平衡量的变化具有良好的鲁棒性。展开更多
High-speed magnitude approximation algorithms for complex vectors are discussed intensively. The performance and the convergence speed of these approximation algorithms are analyzed. For the polygon fitting algorithms...High-speed magnitude approximation algorithms for complex vectors are discussed intensively. The performance and the convergence speed of these approximation algorithms are analyzed. For the polygon fitting algorithms, the approximation formula under the least mean square error criterion is derived. For the iterative algorithms, a modified CORDIC (coordinate rotation digital computer) algorithm is developed. This modified CORDIC algorithm is proved to be with a maximum relative error about one half that of the original CORDIC algorithm. Finally, the effects of the finite register length on these algorithms are also concerned, which shows that 9 to 12-bit coefficients are sufficient for practical applications.展开更多
文摘广义线性结构广泛用于解决共平台系统中同相/正交相位(in-phase/quadrature-phase,IQ)不平衡引起的自干扰镜像信号残留问题。由于在共轭项维度上的拓展,输入信号间的相关性使得用于实现权值向量自适应更新的广义线性复数最小均方(widely linear complex least mean square,WLCLMS)算法出现明显的收敛速度下降。针对这一问题,本文提出了一种二维正交化方法,通过特征值分解实现输入信号在样本延时以及共轭项两个维度上的去相关。同时建立了包括格形预测器、基于梯度的白化器,以及数字对消器三级结构的自适应二维正交化WLCLMS模型,实现了时变环境中自干扰信号的实时跟踪。仿真结果表明,该方法在有效提升收敛速度的同时,有着稳定的稳态误差性能,并且针对系统采样率、滤波器阶数、IQ不平衡量的变化具有良好的鲁棒性。
基金This project was supported by the Natural Science Foundation of Shaanxi Province.
文摘High-speed magnitude approximation algorithms for complex vectors are discussed intensively. The performance and the convergence speed of these approximation algorithms are analyzed. For the polygon fitting algorithms, the approximation formula under the least mean square error criterion is derived. For the iterative algorithms, a modified CORDIC (coordinate rotation digital computer) algorithm is developed. This modified CORDIC algorithm is proved to be with a maximum relative error about one half that of the original CORDIC algorithm. Finally, the effects of the finite register length on these algorithms are also concerned, which shows that 9 to 12-bit coefficients are sufficient for practical applications.
