Hybrid Deep Learning Model for Spectrum Sensing Classification in Cognitive Radio Networks

Spectrum sensing (SS) uses cognitive radio technology at base transceiver stations in order to discover the licensed spectra used by Primary Users (PUs). This enables the SUs to make wireless network transmissions without causing interference. Cognitive Radio Networks (CRNs) help in the problem on the restricted radio frequency spectrum that wireless devices depend on. For enhanced classification of SS in CRN, the proposed research focused on using hybrid Deep Learning (DL) model by integrating Convolutional Neural Network (CNN) and Long Short Term Memory (LSTM). Initially, the data is generated through eight modulation types considering the varied oversampling rate. The spatial features of signal are extracted through CNN and temporal features are identified by LSTM. the Cooperative Spectrum Sensing (CSS) algorithm is integrated increase the reliability of spectrum detection. The hyper parameters of the trained model are fine tuned for enhanced classification of SS. Experimental results showed that the proposed LSTM-CNN-CSS achieved sensing accuracy at 98.34%, precision at 95.62%, recall at 96.77% when compared to the traditional SS classification models such as Ensemble Convolutional Recurrent Neural Network (ECRNN) and LSTM-Extreme Learning Machine (LSTM-ELM).