EXPLORING UNDERMODELING IN COMBINATIONS OF ADAPTIVE FILTERS

Although combinations of filters improve the performance of individual adaptive filters, they increase the computational cost by operating multiple simultaneous filters. In this paper, we propose the undermodeling of the fast filter in a combination to reduce complexity and also obtain improved performance. As an inadequate design of the undermodeled fast filter may affect the combination performance, we describe how the length and the step-size of this filter should be designed, based on analytical models of transient and steady-state performance. We present simulations that evidence the improvement achieved by using this method, also including nonstationary scenarios as well as the application of the proposed method for OFDM wireless communications.