A non-linear local active noise control with integrated virtual sensing using an adaptive hybrid spline filter

Virtual sensing technique has been a promising means for local active noise control (ANC) to surpass the design constraint of error microphone placement at a desired location. Most existing systems which use virtual sensing employ an finite impulse response (FIR) filter based adaptive filter for achieving noise cancellation. However, the effectiveness of such ANC systems deteriorate in the presence of non-linearities in the system. In this paper, an attempt has been made to develop a class of non-linear ANC systems, with integrated virtual sensing, based on adaptive spline filters. These include a Wiener model, a Hammerstein model and a sandwich Hammerstein-Wiener based hybrid model of non-linear ANC. The learning algorithms for the proposed ANC schemes have been developed and it was observed that the adaptive nature of the activation functions allow the ANC system to shift from linear to non-linear and vice versa depending on the control scenario. The simulation results demonstrate that the hybrid spline based ANC system provides an improved noise reduction performance in non-linear scenarios in comparison with other ANC schemes.