Missing Spectrum-Data Recovery in Cognitive Radio Networks Using Piecewise Constant Nonnegative Matrix Factorization

TL;DR

This paper introduces a novel missing spectrum data recovery method for cognitive radio networks using a piecewise constant nonnegative matrix factorization approach, effectively estimating missing data despite noise and fading.

Contribution

It develops a new NMF-based technique with piecewise constant activation coefficients and solves it using Majorization-Minimization, improving spectrum data recovery accuracy.

Findings

Accurately estimates missing spectrum data in noisy conditions

Effective in presence of fading effects

Outperforms existing methods in simulations

Abstract

In this paper, we propose a missing spectrum data recovery technique for cognitive radio (CR) networks using Nonnegative Matrix Factorization (NMF). It is shown that the spectrum measurements collected from secondary users (SUs) can be factorized as product of a channel gain matrix times an activation matrix. Then, an NMF method with piecewise constant activation coefficients is introduced to analyze the measurements and estimate the missing spectrum data. The proposed optimization problem is solved by a Majorization-Minimization technique. The numerical simulation verifies that the proposed technique is able to accurately estimate the missing spectrum data in the presence of noise and fading.