Hybrid QPSO-CNN framework for multi-objective notch band selection in cognitive radio networks

Abstract

Abstract Cognitive radio (CR) systems require dynamic spectrum management and efficient interference mitigation through reconfigurable notch band antennas. This paper proposes a hybrid framework combining Quantum-Inspired Particle Swarm Optimization (QPSO) with Deep Convolutional Neural Networks (CNN) for intelligent notch band switching in sub-6 GHz cognitive radio systems. The CNN performs real-time interference type classification from spectrum measurements, while QPSO optimizes multi-objective notch band configuration considering spectrum efficiency, interference suppression ratio, and switching energy consumption. The complete system is implemented and validated in MATLAB/Simulink using Communications Toolbox and RF Blockset. Simulations using the Kaggle Cognitive Radio Spectrum Sensing Dataset comprising 1000 spectrum measurements across four frequency bands (1.8, 2.4, 3.5, and 5.0 GHz), augmented to 5064 training samples, demonstrate 34.7 ± 4.2% faster convergence than classical PSO (two-sample Welch’s t-test, p < 0.01, n = 500 Monte Carlo trials), 22.3 ± 3.1% improvement in spectrum efficiency under high interference conditions ( p < 0.001), and 41.2 ± 5.8% better interference suppression compared to genetic algorithm-based approaches ( p < 0.001). The proposed framework achieves mean decision latency of 12.02 ± 1.3 ms in software simulation, which is below typical coherence times in mobile channels. The framework is validated for sub-6 GHz cognitive radio applications; extension to full UWB (3.1–10.6 GHz) band requires additional training data collection for frequencies above 6 GHz. Hardware validation on SDR platforms such as USRP B210 is identified as a necessary future step to confirm real-time feasibility beyond simulation.