In most of the existing approaches for neutralizing the effects of acoustic feedback in feed-forward active noise control (FF-ANC) systems, three adaptive filters have been employed. Besides the noise control and feedback neutralization filters (FBNF), a third disturbance removal filter has been exploited to obtain suitable signals for (apposite) adaptation of the other two (adaptive filters). In this paper, we develop an effective acoustic feedback neutralization method which does not require any supplementary third adaptive filter. The proposed method incorporates a probe signal with a time-varying gain and a novel variable step-size strategy for adaptation of the FBNF. During the transient state, the FBNF is excited by a large variance probe signal and a large valued step-size. This enhances the convergence rate of the FBNF. As the FF-ANC system converges, the level of probe noise is decreased to achieve good residual error attenuation performance (at steady-state). Furthermore, the step-size is also reduced to a small value to improve the steady-state performance of the acoustic FBNF. Simulations results show that the proposed two adaptive filters-based method outperforms the existing methods for broad-band reference noise sources and yields better steady-state residual error attenuation performance for narrow-band noise signals.