A hybrid ANC system with a sinusoidal noise canceller equipped with a variable step-size LMS algorithm

If the primary and reference noise signals contain not only wideband but also narrowband components, the conventional broadband active noise control (BANC) and narrowband ANC (NANC) systems may lose their noise reduction merits. Recently, a hybrid ANC system has been proposed to tackle this problem, where a sinusoidal noise canceller (SNC) is introduced to improve the noise environment for the BANC subsystem that follows to perform effectively. The SNC uses an LMS algorithm that has a fixed step size or a variable one of deterministic nature. However, the SNC works inadequately when the noise environment is nonstationary. In this paper, the LMS algorithm used in the SNC is replaced by one with variable step sizes that are of stochastic nature and change according to both the SNC error and the reference signals. This variable step-size LMS based SNC enjoys improved estimation and tracking capabilities as compared to its previous version. The only sacrifice to make is a little increase of implementation cost. Extensive simulations for numerous scenarios are conducted to confirm the effectiveness of the modified hybrid ANC system.