Assessing the acoustic feedback control performance of adaptive feedback cancellation in sound reinforcement systems

Adaptive feedback cancellation (AFC) is considered to be a promising solution to the acoustic feedback problem in sound reinforcement systems. A fundamental problem in AFC is related to the closed-loop nature of a sound reinforcement system, which results in a considerable signal correlation between the far-end and near-end signal. To avoid a biased and slowly converging feedback path estimation, the AFC approach is usually realized by combining an adaptive filter with a decorrelation method. In the AFC literature, decorrelation methods have only been evaluated w.r.t. the resulting adaptive filter misadjustment, and moreover, few results are available concerning the proper choice of the decorrelation parameters. In this paper, results of a comparative evaluation of existing decorrelation methods are reported, in terms of two measures that actually determine the acoustic feedback control performance, namely the maximum stable gain (MSG) increase and the sound quality. It appears that the choice of the decorrelation method and its parameters has a profound influence on these performance measures. Moreover, when decorrelation is applied in the closed signal loop, a trade-off between the resulting MSG increase and sound quality is unavoidable.