Adaptive active noise control
Yang, Xiao Hua
Date of Issue1999
School of Electrical and Electronic Engineering
In this thesis two new adaptive active noise control algorithms using frequency weighting technique are developed. One is based on the filtered-X least-mean-square (FXLMS) algorithm and the other on the filtered-X normalized least-mean-square (FXNLMS) algorithm. They are referred to as the FXLMS++ and FXNLMS++ respectively in the thesis and are motivated by the fact that the FXLMS and FXNLMS algorithms may be easily affected by the random noise which is added to the active noise control system to implement an on-line secondary path identification. Observe that noises to be attenuated in active noise control sys-tems usually are either narrowband signals (for example, industrial fans or trans-formers) or broadband signals with known frequency range. The two proposed algorithms, the FXLMS+-1- and FXNLMS++ algorithms, overcome the drawback of the FXLMS and FXNLMS algorithms by introducing frequency weighting over the frequency range of interest. The frequency weighting allows the algorithms to focus on the noises we want to attenuate while filtering out the frequency com-ponents which are imparted by the random noise added to implement the on-line secondary path identification. Simulations for both a single-input single-output (SISO) system and an multi-input multi-output (MIMO) system are carried out using the four algorithms, namely, the FXLMS, the FXNLMS, the FXLMS++ and the FXNLMS++ algorithms. The results clearly indicate that the proposed algo-rithms, especially the FXNLMS+-1- algorithm, enjoy fast convergence and better atenuation of noises. Two testbeds are then designed to test the applicability of the algorithms. The experimental results further verify our findings in simulations, i.e. the FXNLMS++ algorithm gives the best result for both the SISO and MIMO active noise control.
DRNTU::Engineering::Electrical and electronic engineering::Computer hardware, software and systems