Hybrid Spectrum Sensing Using Neural Network–Based MF and ED for Enhanced Detection in Rayleigh Channel

Abstract

Spectrum sensing (SS) is an integral part of cognitive radio systems, allowing for dynamic spectrum access and efficient exploitation of scarce spectral resources. Classic spectrum sensing methods, such as matched filters (MFs) and energy detections (EDs), usually fail in low-SNR and interference-rich scenarios, with poor detection performance and suboptimal spectrum usage. This work proposes a hybrid spectrum sensing approach that combines the neural network (NN)-based MF and ED to address these limitations. The NNs act as an intelligent signal processor that uses its ability to learn and adapt to different channel conditions to enhance signal detection in low-SNR environments. The proposed framework combines the accuracy of MF with the adaptability of ED, guided by a NN to improve decision-making accuracy. Extensive simulations demonstrate that the method achieves significant improvements in detection accuracy, false alarm reduction, and spectrum hole identification compared to traditional approaches. Furthermore, the capability of the NN to mitigate noise and interference results in enhanced bit error rate (BER) performance, ensuring reliable communication. The paper assesses the system performance in terms of the key metrics, BER, probability of detection (Pd), and probability of false alarm (Pfa), power spectral density (PSD), and capacity, thereby indicating robustness toward dynamic and noisy environments. The results, therefore, open up a potential for defining spectrum sensing by NNs as a scalable, adaptive, and efficient solution for future wireless communication systems in applications such as IoT, 5G, and next-generation cognitive radios.