A nonlinear active noise control scheme with on-line model structure selection

In nonlinear active noise control (ANC) applications the on-line tuning of the parameters of the nonlinear control filter is not sufficient to guarantee the required model accuracy, and a suitable model structure adaptation scheme must be included. However, the ANC setting configures an indirect model identification problem which does not give direct access to the target output signal for the filter model. As such, the filter adaptation problem cannot be solved with the linear regression tools usually employed for model selection. A modified ANC scheme is here proposed, where the controller adaptation loop is reconfigured as a direct identification problem, to allow for model selection, and an auxiliary adaptation loop is introduced to compensate for the error resulting from the scheme modification. Some simulation examples are reported to show the algorithm effectiveness.