Machine-learning-based spectrum sensing enhancement for software-defined radio applications

The Software-defined radio (SDR) technology is considered as a promising solution to address the issue of spectrum scarcity by providing a high level of flexibility and configurability to wireless communication systems. This technology, widely used in cognitive radio (CR) networks, aims at optimizing the spectrum efficiency. However, to reach this objective, a new conception of spectrum sensing methods needs to be adopted. Indeed, in the emerging wireless communications systems, telecom researchers face the challenge of detecting the presence of energy with a very low signal-to-noise ratio (SNR) in a very complex environment. For this purpose, for complex environments such as CR networks, machine-learning (ML) algorithms seem to be a suitable solution allowing a smart spectrum-sensing scheme. This scheme enables the increase of the signal detection probability and the decrease the false detection probability of signal detection. In this paper, four supervised ML models are compared: Naïve Bayes classifier, Support Vector Machine, Gradient Boosting Machine and Distributed Random Forest. Furthermore, a principal component analysis (PCA) is performed to reduce the dimensionality of the complex dataset while preserving the useful variability. Their performance is subsequently evaluated by computing their receiver operating characteristic (ROC) curves. The entire algorithms are then implemented and demonstrated using the GNU (GNU’s Not Unix) Radio platform and an SDR-based electronic card, HackRF One.