基于凸优化的频域自适应滤波方法

A Frequency Domain Adaptive Filter Method Based on Convex Optimization

传统的自适应滤波方法均是采用时域算法,对具有多个中心频率的有限带宽含高斯噪声信号的降噪效果均不理想。为了解决有限带宽含高斯噪声信号的降噪问题,本文基于凸优化方法设计出一种多个带通Wiener滤波器并联组合的频域自适应滤波器,采用拉伊达准则确定有限带宽信号的中心频率,给出多个带通Wiener滤波器的频域参数的优化目标函数,并且通过仿真试验和与小波降噪效果对比说明了本文所提出方法的有效性,验证了该方法对信号具有较强的自适应能力,能够很好的消除实测信号中的高斯噪声。最后给出工程实例验证了方法的有效性和实用性。

Traditional adaptive filter methods are based on time domain analysis, which have poor performance for signals with multi central frequency, limited bandwidth and white noises. Based on convex optimization method, we design a adaptive filter to denose the limited bandwidths signal with center frequency, and the filter is consisted of multi parallel Wiener filters which are band pass filters. We established the optimization objective function of frequency domain parameter and get the center frequency with the Pauta criterion.Finally, we apply this method to a simulation experiment and an engineering example for demonstrating its efficiency in de-nosing.