Improved spectrum prediction model for cognitive radio networks using hybrid deep learning technique

Abstract

Cognitive Radio (CR) technology has been highlighted as one of the most likely answers to the issue of spectrum shortage with the rise of fifth generation and beyond communication. Secondary users (SUs) in cognitive radio networks (CRN) must continuously monitor the spectrum to forecast channel occupancy by primary users (PUs) based on fundamental factors, such as location, time, and RF band. A hybrid deep learning model called LSTM-MLP (Long Short-Term Memory-Multilayer Perceptron) is proposed to improve idle channel prediction probability thus reducing the overall sensing time by cognitive users during spectrum sensing. Performance evaluation for the proposed model is done in terms of prediction error and efficiency, the GSM-900 spectrum dataset demonstrates that LSTM-MLP performs better in terms of improved prediction accuracy compared to existing state-of-art prediction techniques.