在线参数辨识的脉冲噪声有源控制

Active Control of Impulsive Noise Based on On-line Parameter Identification

有源噪声控制是一种主动控制方法,目前已广泛应用于对高斯分布噪声进行衰减。但是传统的用于控制噪声的自适应算法不再适用大多数服从非高斯分布的脉冲噪声,主要原因是这种脉冲噪声没有有限的二阶统计量。在经典的Filter-x LMS算法的基础上提出两种适用于服从非高斯分布尖峰脉冲噪声情况下的在线参数辨识方法,一种是利用在线参数辨识方法对服从SαS稳定分布的脉冲噪声进行特征指数的估计,进而实现降噪目的的FxLMPest和FxLMADadapt算法;另一种是在Sun等人提出的SKM和AM算法基础上利用在线递归过程实现对幅度阈值估计的BDP算法。这两种算法均不需要获得脉冲噪声的特征指数和阈值的先验信息,仿真分析结果表明这两种算法能有效抑制脉冲噪声,并且其鲁棒性明显好于Filter-x LMS算法。

Active noise control is a kind of active control methods and has been widely used to attenuate the noise which yields Gaussian distribution. But for most of non-Ganssian impulsive noises, the traditional adaptive algorithms are not appropriate for controlling the impulsive noise. The main reason is that there is no finite second-order moment for the impulsive noise. In this paper, two kinds of on-line parameter identification algorithms based on the classical Filter-x LMS algorithm were presented which are appropriate for non-Gaussian impulsive noise. One is FxLMPest and FxLMADadapt algorithms based on on-line parameters identification to estimate characteristics exponent of impulsive noise with an S a S distribution, and the other is BDP algorithm based on SKM and AM algorithms presented by Sun et al, which uses a simple on-line recursive procedure to estimate amplitude thresholds. Both methods do not need prior information of characteristics exponent and amplitude thresholds of impulsive noise. And the simulation results also show that the two methods can effectively suppress the impulsive noise, and its robust performance is significantly better than Filter-x LMS algorithm.