The aim of the project is related to active control of in-car noise. Basically, an artificially generated sound is superimposed on the unwanted interior noise in order to cancel it out. The canceling signal is obtained by suitable detection and processing of the interior noise so that the unwanted sound perceived by the human observer is considerably reduced. The core of the system consists of an ensemble of “experts' that are specialized in modeling the in-car noise for predefined engine rotation intervals. These are implemented by artificial neural networks, due to their well-known approximation capabilities. Tracking capabilities of the changes occurring in the environment are provided by adaptive weighting of the experts outputs. This action is driven by a discriminator that is able to distinguish between useful sounds (voice, radio, alarm signals) and unwanted noise. Experimental results show the efficiency of the proposed method, indicating significant noise level reduction in a fairly broad frequency range.