基于自适应滤波的多模噪声抑制研究

Research of Multi-Mode Noise Suppressing Based on Adaptive Filter

多数研究都是将噪声当作高斯噪声处理,但实际上在多模噪声(整体上属于非高斯噪声)背景下,信号受损严重。用传统的LMS自适应算法不能很好的抑制噪声,格里菲斯LMS算法收敛速度较慢。提出一种将格里菲斯LMS算法(GLMS)与LMS-Newton算法相结合的GLMS-Newton算法,并给予了改进。在改进中,不仅采用了基于互相关变步长因子,而且引入了基于输入信号与期望信号的互相关的梯度算子。仿真表明,在收敛速度、稳态误差等方面都有了较大改善,能够很好的抑制多模噪声,提取出有用信号。

Noises are treated as a Gaussian noises in most researches,but actually in multi-modal noises（the whole belongs to Gaussian noise） background,signals are severely damaged.Traditional LMS adaptive algorithm cannot better reduce the noises and Griffith LMS algorithm converges slowly.This paper proposed an GLMS-Newton algorithm which combined GLMS algorithm with LMS-Newton algorithm,and the improvement was given.The algorithm not only adopted the variable step size factor,but also introduced on the gradient operator based on the cross-correlation of the input signals and the desired signals.The simulation show that the algorithm can improve the convergence speed,reduce the steady-state error,and effectively suppress the multi-modal noise.