SVM-MUSIC algorithm for spectrum sensing in cognitive radio ad-hoc networks

TitleSVM-MUSIC algorithm for spectrum sensing in cognitive radio ad-hoc networks
Publication TypeConference Paper
Year of Publication2017
AuthorsBarrak, S. El, A. Lyhyaoui, A. El Gonnouni, A. Puliafito, and S. Serrano
Conference NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Conference LocationCham
ISBN Number9783319679099} pages = {161–170
KeywordsCognitive radio ad hoc networks (CRAHNs), Computer Science (all), Minimum variance distortionless response (MVDR), Multiple signal characterization (MUSIC), Spectrum sensing (SS), Support vector machines (SVM), Theoretical Computer Science

Adopting accurate and efficient spectrum sensing policy is crucial in allowing cognitive radio users to be aware of the surrounding parameters related to the radio environment characteristics. Especially, in Ad hoc networks scenario, where dynamic spectrum access is highly required, since the most of the spectrum is already assigned statistically, and the unlicensed bands are becoming overcrowded. This is due to the multiplicity of wireless communication technologies that operate in those bands, and the increasing number of connected devices. In this paper, a spectrum sensing algorithm that combines the Support Vector Machines (SVM) supervised learning technique with the Multiple Signal Characterization (MUSIC) subspace method is used. Our ultimate objective is detecting the presence of primary users (technology signals) in the band of interest. The node’s receivers which make up the network collect samples from the radio environment, estimate the number of primary user signals and the corresponding carrier frequencies. Simulations are conducted to demonstrate the efficiency of the proposed SVM based algorithm in detecting the presence of primary users based on lost-detection and false alarm probabilities evaluation.