A multi-resolution filtered-x LMS algorithm based on discrete wavelet transform for active noise control
A multi-resolution filtered-x LMS algorithm based on discrete wavelet transform for active noise control
复制标题
DOI：
10.1016/j.ymssp.2015.05.024
复制
发表时间：
2016
影响因子：
8.4
通讯作者：
Z. Qiu;C.-M. Lee;Z. H. Xu;L. Sui
中科院分区：
文献类型：
--
作者：
Z. Qiu;C.-M. Lee;Z. H. Xu;L. Sui
我们开发了一种基于离散小波变换 (DWT) 的新型主动控制算法，适用于稳态和非稳态噪声控制。首先，引入Mallat金字塔算法来实现DWT，该算法可以将参考信号分解为多个具有多分辨率的子带，并提供完美的重建（PR）过程。为了减少 DWT 引入的额外计算复杂度，提出了一种有效的策略，使用快速傅立叶变换 (FFT) 更新频域 Deepthi B.B 中的自适应滤波器系数。基于参考噪声源，采用“Haar”小波，通过将噪声信号分解为两个子带（3频带），所提出的基于DWT-FFT的FXLMS（DWT-FFT-FXLMS）算法与时域滤波x最小均方（TD-FXLMS）算法相比，大大降低了复杂度并具有更好的收敛性能。由于小波分析突出的时频特性，所提出的DWT-FFT-FXLMS算法可以有效地消除平稳和非平稳噪声，而频域FXLMS（FD-FXLMS）算法无法达到这一点。
We have developed a new active control algorithm based on discrete wavelet transform (DWT) for both stationary and non-stationary noise control. First, the Mallat pyramidal algorithm is introduced to implement the DWT, which can decompose the reference signal into several sub-bands with multi-resolution and provides a perfect reconstruction (PR) procedure. To reduce the extra computational complexity introduced by DWT, an efficient strategy is proposed that updates the adaptive filter coefficients in the frequency domainDeepthi B.B using a fast Fourier transform (FFT). Based on the reference noise source, a ‘Haar’ wavelet is employed and by decomposing the noise signal into two sub-band (3-band), the proposed DWT-FFT-based FXLMS (DWT-FFT-FXLMS) algorithm has greatly reduced complexity and a better convergence performance compared to a time domain filtered-x least mean square (TD-FXLMS) algorithm. As a result of the outstanding time-frequency characteristics of wavelet analysis, the proposed DWT-FFT-FXLMS algorithm can effectively cancel both stationary and non-stationary noise, whereas the frequency domain FXLMS (FD-FXLMS) algorithm cannot approach this point.