A recursive least squares algorithm with reduced complexity for digital predistortion linearization

In digital predistortion (DPD) implementation, the computational complexity of coefficients estimation of the DPD model is a key performance metric. Conventional coefficients estimation algorithms, such as least squares (LS), recursive least squares (RLS), and least mean squares (LMS) cannot achieve a fast convergence with little computation. In this paper, we propose an RLS algorithm with reduced complexity by introducing orthonormal polynomial basis functions. The proposed algorithm is as simple as LMS algorithm yet as efficient as RLS algorithm. Simulation results validate our analysis.