Improving the performance of an adaptive filtering structure by applying the generalised structural subband decomposition

In this paper a method is presented for improving the performance of an adaptive filtering structure (AFS) by applying the generalised structural subband decomposition (GSSD). The method is applied to a maximally decimated (MD) structure, which has been recently proposed in the literature, and a way is developed for identifying the generalised polyphase components (GPC) of the polyphase components (PC) of the system under modelling. A detailed convergence analysis is presented and experimental results are given that demonstrate the improvement in the convergence rate (CR) the proposed solution offers for coloured input processes. Particular attention is paid to the white input case. Specifically, a new result related to cosine modulated (CM) filter banks (FB) is given and used to show that the introduced application of GSSD does not affect the CR when the input is white. Finally, the proposed modifications do not practically increase the computational complexity and thus the new method is a viable solution for practical applications.