信号插值对归一化变步长FxLMS算法的影响研究

Effect of signal interpolation on the performance of normalized variable step size FxLMS algorithm

自适应滤波算法是主动控制系统的核心部分,为进一步优化归一化变步长滤波x-最小均方(NVSS-FxLMS)算法,本研究引入信号插值概念,构建结合信号插值与归一化变步长理念的INVSS-FxLMS算法,以进一步改善控制效果及算法收敛速度,并探究了信号插值数量对算法性能的影响。仿真结果表明,信号插值可改善NVSS-FxLMS算法的控制效果及收敛速度,且插值数量存在最佳范围,当插值数量为10左右时,控制效果及收敛速度较好,总声压级/加速度级降低率达18%,而计算时间仅增加约9.6%,为NVSS-FxLMS算法的发展提供了新思路,并为主动控制系统的改进奠定基础。

The adaptive filter algorithm is the core of active control systems(ACS).To further improve the performance of the normalized variable step size filter-x least mean square(NVSS-FxLMS)algorithm,the signal interpolation concept is introduced in this research,and the algorithm combining the signal interpolation with normalized variable step size concept(INVSS-FxLMS algorithm)is proposed to improve the control performance and convergence rate of the NVSS-FxLMS algorithm,then the effect of the number of signal interpolation on the performance of the algorithm is explored.The results show that the signal interpolation can improve the control performance and convergence rate of the NVSS-FxLMS algorithm.Besides,the optimal range of the number of signal interpolation exists.When the number is about 10,the control performance and convergence rate of the algorithm is relatively better,and the total sound pressure level or acceleration level in the frequency domain is reduced by about 18%,while the computational time increases by 9.6%.The research provides a new idea for the improvement of the NVSS-FxLMS algorithm,and lays the foundation for the development of the ACS.