A compressed sensing method for pruning concurrent dual-band power amplifier models

This paper presents a compressed sensing (CS) method for pruning concurrent dual-band power amplifier (PA) behavioral models. The method is based on the sparsity assumption for kernel and coefficients of a dual-band PA model. The orthogonal matching pursuit (OMP) algorithm is adopted as a substitution of the traditional least square method to extract the model coefficients. Experimental results show that the similar model accuracy and DPD performance can be achieved by the proposed method, while the model complexity is lower because the number of the kernel and coefficients is reduced.