Theoretical Analysis of Cyclic Frequency Domain Noise and Feature Detection for Cognitive Radio Systems

TL;DR

This paper provides a theoretical analysis of noise distribution in the cyclic frequency domain for cognitive radio spectrum sensing, demonstrating that GEV distribution accurately models noise and improves detection threshold setting.

Contribution

It introduces the use of GEV distribution to model cyclic frequency domain noise and validates this model through maximum likelihood estimation and Monte Carlo simulations.

Findings

GEV distribution accurately models cyclic frequency domain noise

Theoretical ROC curves match simulation results

Enhanced threshold setting for spectrum sensing

Abstract

In cognitive radio systems, cyclostationary feature detection plays an important role in spectrum sensing, especially in low SNR cases. To configure the detection threshold under a certain noise level and a pre-set miss detection probability Pf, it's important to derive the theoretical distribution of the observation variable. In this paper, noise distribution in cyclic frequency domain has been studied and Generalized Extreme Value (GEV) distribution is found to be a precise match. Maximum likelihood estimation is applied to estimate the parameters of GEV. Monte Carlo simulation has been carried out to show that the simulated ROC curve is coincided with the theoretical ROC curve, which proves the efficiency of the theoretical distribution model.