Table 2.
Summary of recent research in cognitive radio spectrum sensing and adaptive modulation.
| Ref | Study | Method/model | Key findings |
|---|---|---|---|
| 20 | Deep learning-based multi neural network architecture for spatial cognitive radio | Multi-domain fusion deep learning architecture | Enhanced signal detection in challenging environmental conditions through multi-domain fusion, improving spectrum sensing (SS) in CR systems |
| 21 | Time-frequency cross fusion network (TFCFN) for spectrum sensing | GRU for temporal feature extraction, CNN for spatial feature fusion in non-Gaussian noise environments | Significantly improved detection performance under non-Gaussian noise, highlighting the effectiveness of combining temporal and spatial features |
| 22 | Assessment of SU mobility effect on SS performance | Real-time adaptive modulation methods and recursive estimation techniques | Demonstrated necessity for real-time adaptive modulation due to mobility-induced fluctuations in signal strength, impacting AMC accuracy |
| 23 | Feature selection in spectrum sensing techniques for 5G CR networks | Re-standardized energy detection and complex feature extraction for ML models | Highlighted the robustness and improved accuracy of complex features over conventional ones, especially at low SNR levels |
| 24 | Recursive estimation for mobile users in cluttered environments | Recursive estimation to mitigate signal strength fluctuation effects | Improved spectrum sensing reliability in dynamic environments affected by multipath fading and shadowing |
| 25 | Deep learning-based AMC with transfer learning on RadioML2016.10b dataset | CNN-based transfer learning model | Achieved improved modulation classification performance across varying SNR levels, demonstrating the promise of CNN architectures for AMC in CR networks |
| 26 | Spectrum mobility optimization in congested CR-LANs | Simple recursive estimator, Kalman and alpha-beta filters for SU tracking | Reduced total transmission time to 4.1827 s and improved end-to-end throughput to 3.67 kbps |
| 27 | Cooperative AMC using higher-order cumulant-based features | Spatially distributed sensor nodes and weight vector-based controllability factor | Improved classification accuracy by 15% at 10 dB SNR compared to traditional higher-order statistics-based methods |
| 28 | Hybrid prediction in multipath and shadowed fading (MSF) environments | Alpha-Beta filtering with Neyman–Pearson type detector | Reduced detection error probability and achieved an average relative error of  , improving primary signal prediction in MSF conditions |