A two-stage adaptive algorithm in the frequency domain for a multichannel feedforward active noise control system

The reference paths between original sources and reference sensors in multichannel feedforward active noise control (ANC) systems are often ignored by most ANC algorithms. Therefore, a two-stage adaptive algorithm in the frequency domain is proposed to deal with several of the more complicated cases, specifically addressing instances when the reference sensors must be placed far away from the noise sources. This two-stage system includes a blind pre-processing system cascaded with a traditional MFxLMS system; both systems operate in the frequency domain. The new algorithm can eliminate the effect of reference paths by performing blind pre-processing independently. The MFxLMS algorithm in the following stage can achieve faster convergence than the traditional algorithm without blind pre-processing. The computational complexity of the proposed algorithm is analysed and a numerical simulation using impulse responses measured in a real reverberant room is performed to verify the convergence performance of the proposed algorithm.