Binary is Good: A Binary Inference Framework for Primary User Separation in Cognitive Radio Networks

TL;DR

This paper introduces a binary inference framework for separating primary users in cognitive radio networks, enabling accurate detection without prior activity knowledge and efficient representation of multiple sources.

Contribution

It proposes a novel binary inference algorithm for PU separation that operates without prior knowledge and efficiently encodes multiple sources with logarithmic-length vectors.

Findings

High inference accuracy achieved without prior knowledge

Effective representation of multiple binary sources

Algorithm outperforms existing methods in simulations

Abstract

Primary users (PU) separation concerns with the issues of distinguishing and characterizing primary users in cognitive radio (CR) networks. We argue the need for PU separation in the context of collaborative spectrum sensing and monitor selection. In this paper, we model the observations of monitors as boolean OR mixtures of underlying binary latency sources for PUs, and devise a novel binary inference algorithm for PU separation. Simulation results show that without prior knowledge regarding PUs' activities, the algorithm achieves high inference accuracy. An interesting implication of the proposed algorithm is the ability to effectively represent n independent binary sources via (correlated) binary vectors of logarithmic length.