Structured compressed-sensing for volterra series models

Abstract

This paper presents a simple and efficient algorithm for improving the sparse modeling of wireless communications nonlinear devices. This technique consists on a modified greedy algorithm that sorts the regressors of the model by its importance and a maximum likelihood method for the selection of the optimum number of coefficients. This approach has been applied to obtain a model with a reduced number of parameters while upholding its performance capabilities for a commercial power amplifier (PA) driven with an orthogonal frequency division multiplexing signal, following the LTE standard. Results show that the presented algorithm achieves the same level of performance than a reference compressed-sensing algorithm while providing a further reduction in the number of coefficients.

Digital predistortion of power amplifiers using structured compressed‐sensing Volterra series

“…These strategies can be either ad-hoc, which include a subset of the FV such as the generalised memory polynomial (GMP) or dynamic deviation reduction among others, or based on information theory, which do not include the information of the intrinsic structure of the model [3]. Structural information based on the algorithm in [4] was incorporated by the authors in [5]. In this Letter, we show the structural pruning of Volterra series and validate the method in the DPD application, obtaining a reduced complexity model while keeping the level of performance.…”