Spectrum Sensing in Cognitive Radio using Random Forest with Wavelet and Empirical Mode Decomposition

Abstract

Employing a stochastic approach to cognitive radio, this study meticulously assesses the efficacy of cooperative spectrum sensing strategies, specifically leveraging the Random Forest machine learning paradigm.  Comparative analysis is undertaken across two distinct feature extraction methodologies: wavelet transformation and empirical mode decomposition.  The primary user signal is modelled as a 5G-compliant orthogonal frequency-division multiplexing (OFDM) waveform. The cooperative sensing framework employs a majority voting fusion rule. Simulation results demonstrate the effectiveness of both feature extraction techniques combined with Random Forest for robust spectrum sensing, highlighting their comparative performance under varying noise conditions.