Parallel implementations with low-complexity of rotation-based adaptive filters

Abstract

This paper propose a new parallel implementation of some rotation-based adaptive filters (Bhouri, M, 2000). These filters are characterized by a robust behavior to input signal correlation (Bhouri, M, et al, 1998) and good numerical properties. However, their implementations have reduced complexities. The circuits based on these block-diagonal adaptive algorithms use less computing cells than the systolic circuit of the QR-RLS algorithm. Nevertheless, these new and low-complexity architectures have no longer a pipeline structure.