EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

A Deep Learning Approach to Matched Filter-Based Spectrum Sensing

Md. Tofail Ahmed, Mousumi Haque, Yosuke Sugiura, Tetsuya Shimamura

IEEE Open Journal of the Communications Society · 2026 · 7 · 8633-8645

DOI: 10.1109/ojcoms.2026.3714491 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Open Journal of the Communications Society, 2026. Verfasst von Md. Tofail Ahmed und 3 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In cognitive radio (CR) networks, matched filter detection (MFD) is widely recognized for its high detection performance compared to other spectrum sensing methods, such as energy detection (ED). The MFD requires both knowledge of the primary user (PU) signals and the noise statistics at the secondary user (SU), which is difficult to obtain in practice, leading to increased computational complexity. In practical scenarios, transmitted PU signal information is generally more difficult to obtain than noise information. Moreover, conventional methods suffer from threshold selection issues, as a high threshold increases the missed detection probability, whereas a low threshold leads to more false alarms. To address these limitations, this study proposes a semi-blind matched filter detection (SBMFD) method based on deep learning (DL). The proposed SBMFD framework requires only noise information, not knowledge of the PU signal, making it easier to implement in the real world while retaining the enhanced PU detection capability of MFD. In simulation, we consider the PU signals modulated with quadrature amplitude modulation (QAM) that are transmitted over a Rayleigh fading channel and corrupted by additive white Gaussian noise (AWGN). The performance evaluation results demonstrate that the SBMFD method outperforms the conventional MFD method, the SBMFD method without noise information (WSBMFD), as well as state-of-the-art DL-based spectrum sensing methods. Specifically, the proposed method provides a probability of detection ( $P_{d}$ ) of 97.76% at a low SNR of −24 dB.

Zitieren

APA

Ahmed, M. T., Haque, M., Sugiura, Y., & Shimamura, T. (2026). A Deep Learning Approach to Matched Filter-Based Spectrum Sensing. IEEE Open Journal of the Communications Society, 7, 8633-8645. https://doi.org/10.1109/ojcoms.2026.3714491

Vancouver

Ahmed MT, Haque M, Sugiura Y, Shimamura T. A Deep Learning Approach to Matched Filter-Based Spectrum Sensing. IEEE Open Journal of the Communications Society. 2026;7:8633-8645. doi:10.1109/ojcoms.2026.3714491

BibTeX
@article{Ahmed2026,
  author = {Ahmed, Md. Tofail and Haque, Mousumi and Sugiura, Yosuke and Shimamura, Tetsuya},
  title = {A Deep Learning Approach to Matched Filter-Based Spectrum Sensing},
  journal = {IEEE Open Journal of the Communications Society},
  year = {2026},
  volume = {7},
  pages = {8633--8645},
  doi = {10.1109/ojcoms.2026.3714491}
}
Kostenlos starten →