Spectral decision in cognitive radio networks based on deep learning

Abstract

Listen

Translate article

Cognitive radio networks (CRN) have gained great relevance in the efficient use of the radio spectrum, and one of the key aspects of this technology is the spectral decision. The performance of secondary user communication depends largely on the intelligent choice of an appropriate spectral opportunity. The purpose of this research is to propose and assess the performance of a spectral decision model for CRN based on the Deep Learning technique. To achieve this, a classifier was adapted through the feature extraction technique that identifies three levels of traffic (high, medium and low) in a spectral occupation experimental power matrix that models the primary user. The extraction of features is done by Deep Learning and the process of classifying the successful set of features is done by a Support Vector Machine (SVM). These were used along with five evaluation metrics—total handoffs, failed handoffs, bandwidth, delay and throughput—to measure the performance of the proposed spectral decision model based on the Deep Learning technique, and to compare the results with the Multi-Criteria Optimization and Compromise Solution (VIKOR), Technique for Order Preference by Similarity to Ideal Solution (TOPSIS), and Simple Additive Weighting (SAW). This work presents five contributions: incorporation of the real behavior of licensed users, implementation of performance metrics for spectral mobility, proposal of an RGB conversion algorithm based on the threshold level, feedback in the classifier and a methodology based on priorities and scores to establish the channels with the highest availability. The results of this evaluation show that the proposed model has a better performance in the five metrics compared to the other techniques.