Abstract
Cooperative spectrum sensing (CSS) plays a vital role in cognitive radio networks (CRNs). CSS enables efficient spectrum utilization and improves communication reliability through dynamic detection of underutilized frequency bands. However, the methods which evolved earlier face limitations like reduced detection accuracy and high false alarm rates in low Signal-to-Noise Ratio (SNR) conditions. Efficient allocation of spectrum resources in cognitive radio networks is affected by unreliable sensing at low signal-to-noise ratios, which leads to missed detections and false alarms. To overcome these challenges a novel optimized deep learning model is presented in this research work which provides reliable spectrum detection under noisy environments. The proposed Optimized Multi-Scale Graph Neural Network with Attention Mechanism (OMSGNNA) utilizes graph-based data representation and provides multi-scale feature extraction. The attention mechanism effectively fuses spatial and temporal information which enhances the performances. Additionally, an Adaptive Butterfly Optimization with Lévy Flights (ABO-LF) is incorporated to fine tune the parameters of proposed model. The proposed model experiments which are conducted using benchmark RadioML2016.10b dataset includes I/Q samples from 11 modulation schemes across SNR levels ranging from - 20 dB to 18 dB. The performance evaluation exhibits better performance of proposed OMSGNNA in terms of 98% accuracy at high SNR with better precision, recall, and F1-score. Comparative analysis reveals that the proposed OMSGNNA outperforms traditional deep learning models even at low SNR conditions.