Case study of TV spectrum sensing model based on machine learning techniques

Abstract

Spectrum sensing is an essential component in cognitive radios (CR). Machine learning (ML) algorithms are powerful techniques for designing a promising spectrum sensing model. In this work, the supervised ML algorithms, support vector machine (SVM), k-nearest neighbor (kNN), and decision tree (DT) are applied to detect the existence of primary users (PU) over the TV band. Moreover, the Principal Component Analysis (PCA) is incorporated to speed up the learning of the classifiers. Furthermore, the ensemble classification-based approach is employed to enhance the classifier predictivity and performance. Simulation results have shown that the highest performance is achieved by the ensemble classifier. Moreover, simulation results have shown that employing PCA reduces the duration of training while maintaining the performance.