可變步長演算法於有無干擾之主動噪音控制系統的研究

Abstract

主動噪音控制系統在實際應用中，環境的改變將會影響主動噪音控制的性能。為了處理環境改變的問題，本文提出了兩種可變步長正規化LMS演算法，一種是以殘留誤差直接調整步長的型式，另一種是以適應性方法更新步長的型式。並以一個已建立的決定論方法進一步分析可變步長正規化演算法，使得分析出的條件在主動噪音控制中能確保演算法可以收斂。而在電腦模擬中，我們考慮四種情況，包含改變主噪音源之振幅、改變主噪音源之頻率、改變第一路徑或改變第二路徑的情況。由模擬結果可得知，應用本文所提出的可變步長演算法於這些環境改變的情況之下，可以有效保持主動噪音控制之消音性能，並實現演算法的可行性。

In practical active noise control (ANC) systems, environmental changes will affect ANC performance. This study proposes two kinds of variable step sizes (VSSs) for the normalized LMS (NLMS) algorithms to deal with the problems caused by environmental changes. One uses a residual error directly to tune step size, while another uses an adaptive method to update step size. A deterministic method is then established to analyze the VSS-based NLMS algorithms such that conditions to ensure convergence of the algorithms in ANC systems can be derived. Four kinds of environmental changes, including change in amplitudes of the primary noise, change in frequencies of the primary noise, change of the primary path, and change of the secondary path, are considered in computer simulations. Simulation results show that the VSS-based algorithms can maintain effective ANC performance under these environmental changes and thus support the feasibility of the proposed VSSs.