A PSO-based weighting method to enhance machine learning techniques for cooperative spectrum sensing in CR networks

A Comprehensive Survey on Spectrum Sensing in Cognitive Radio Networks: Recent Advances, New Challenges, and Future Research Directions

“…The state ℋ0 corresponds primary user absence and state ℋ1 corresponds primary user presence. For the sensing decision, several of the previously mentioned spectrum sensing techniques can be used, including energy detection [5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20], cyclostationary detection [21,22,23,24,25,26,27], matched filter detection [28,29,30,31], covariance-based detection [32,33,34,35,36,37,38,39], and machine-learning based detection [40,41,42,43,44,45,46,47,48,49,50,51] which are discussed below. These techniques are often evaluated using the probabilities of false alarm and probability of detection.…”

“…Machine learning has received increasing interest and has found application in many fields due to its ability to apply complex mathematical calculations to analyze and interpret patterns and structures in data, enabling learning, reasoning, and decision making. In the context of cognitive radio networks, several research papers related to machine learning for spectrum sensing have been published [42,43,44,45,46,47,48,49,50,51]. These machine learning-based sensing techniques aim at detecting the availability of frequency channels by formulating the process as a classification problem in which the classifier, supervised or unsupervised, has to decide between two states of each frequency channel: free or occupied.…”

“…Table 2 shows examples of works that apply machine learning for spectrum sensing to detect the availability of frequency channels [42,43,44,45,46,47,48,49,50,51]. For instance, the authors of [42,43] proposed a spectrum sensing model for cognitive radio based on both K-means and support vector machine.…”

An enhanced two‐phase SVM algorithm for cooperative spectrum sensing in cognitive radio networks

“…2 On the one side, CR is a dynamic system with the capability of autonomous learning and adapting to its environment in order to tune its radio parameters. 11 The authors in another study, 12 further improve the aforementioned method by using the idea of classifiers fusion where SVM, KNN, and naïve Bayes classifiers are combined and the labels of energy features are determined based on the optimized weighted vote of each classifier using particle swarm optimization (PSO) algorithm. More specifically, since CR is a highly-reconfigurable software-defined radio, the degree of freedom of wireless system is increased considerably, which adds complexity to the problem.…”

“…Also, two supervised classification techniques, K-nearest-neighbors (KNN) and SVM, in addition to two unsupervised clustering methods, Gaussian mixture model (GMM) and K-means, are applied on energy features for cooperative SS in an online fashion in a previous study. 11 The authors in another study, 12 further improve the aforementioned method by using the idea of classifiers fusion where SVM, KNN, and naïve Bayes classifiers are combined and the labels of energy features are determined based on the optimized weighted vote of each classifier using particle swarm optimization (PSO) algorithm. In a previous study, 13 K-means clustering as an unsupervised learning method has been proposed for cooperative SS in generalized κ-μ fading channels.…”