In this paper, an adaptive robust feedback (FB) active noise control (ANC) system is proposed and applied to the previously developed active noise barrier (ANB) using a hybrid (HB) ANC system, which aims to relieve the indoor noise problem. The FB control is preferred for this small-typed ANB application because only error sensor is required for the FB control system so that a compact controller can be easily realized. In practice, however, the robust stability of the FB controller must be considered. Moreover, the waterbed effect of the FB control will cause noise enhancement at some frequencies out of the target band of control. In order to design a robust controller with limited noise enhancement in real-time control, a modified frequency domain block LMS (modified FDB-LMS, MFDB-LMS) algorithm, which has the reduced computation complexity and the same steady-state performance comparing with the conventional FDB-LMS algorithm, is proposed in this paper, firstly, owing to the easiness to verify the robust stability and the noise enhancement constraints in the frequency domain adaptive process. Then, an internal model control (IMC) typed FB-ANC system exploiting the MFDB-LMS algorithm, into which the constraints to guarantee the robust stability and limit the noise enhancement to within a given value has been integrated, is proposed and applied to the ANB. The effectiveness of the proposed FB-ANC system and the noise attenuation performance of the ANB using the proposed system have been validated by simulations.