The ℓ0-Norm Constraint Coefficient Reusing Least Mean Square Algorithm

Resende, Leonardo; Siqueira, Newton; Diego B. Haddad, Mariane R Petraglia

DOI: 10.14209/sbrt.2019.1570558660

Evento: XXXVII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2019)

Keywords:

Abstract

An adaptive filtering algorithm should present fast convergence and good steady-state behaviour. Both metrics can be enhanced if one makes use of the sparsity (energy concentration in few coefficients) of the involved (unknown) transfer function. Unfortunately, such an approach can suffer from steady-state performance degradation when the signal-to-noise ratio is low, due to an increase of the adaptive weights variance. This work advances a new algorithm that combines the reusing coefficient strategy, which increases the robustness against the additive noise, with a sparsity-aware algorithm, which employs an approximation of the l0-norm in order to penalize non-sparse adaptive vector estimations.

Download