EUVIMEDEUVIMED entdecken
Crossref

Deep Learning-Based Signal Modulation Classification for Cognitive Radio Networks Using CNN-BiLSTM

Ajay Sonkesriya

International Journal of Recent Development in Engineering and Technology · 20.12.2023 · 12 (12) · 20-28

DOI: 10.54380/ijrdet1223_5 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Recent Development in Engineering and Technology, online erschienen am 20.12.2023. Verfasst von Ajay Sonkesriya.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Cognitive radio (CR) technology has emerged as a transformative paradigm for resolving the growing spectrum scarcity problem in modern wireless communication systems. At the core of every cognitive radio implementation is the ability to correctly identify the modulation scheme of an incoming signal, a task known as Automatic Modulation Classification (AMC). Accurate AMC enables the secondary user to make intelligent decisions about spectrum access, interference avoidance, and adaptive transmission without any cooperation from the primary user. Classical AMC methods based on maximum likelihood estimation and hand-crafted statistical features exhibit satisfactory performance only under high signal-to-noise ratio conditions and degrade sharply under multipath fading, frequency offsets, and low SNR regimes common in real cognitive radio environments. This paper proposes a deep learning architecture, CNN-BiLSTM, that integrates a Convolutional Neural Network for local feature extraction with a Bidirectional Long Short-Term Memory network for sequential dependency modelling. The proposed model is trained and evaluated on the widely adopted RadioML 2018.01A benchmark dataset comprising 24 modulation types and 2.5 million signal samples across SNR levels from negative 20 dB to positive 30 dB. Experimental results demonstrate that the proposed CNN-BiLSTM achieves a classification accuracy of 96.8% at SNR of 10 dB and above, surpassing standalone CNN (91.4%), standalone LSTM (88.7%), BiLSTM (90.2%), ResNet (93.6%), and VGG-based architectures (92.1%) by statistically significant margins. The proposed model additionally demonstrates robust performance at challenging SNR levels below 0 dB, where classical approaches fail entirely.

Zitieren

APA

Sonkesriya, A. (2023). Deep Learning-Based Signal Modulation Classification for Cognitive Radio Networks Using CNN-BiLSTM. International Journal of Recent Development in Engineering and Technology, 12(12), 20-28. https://doi.org/10.54380/ijrdet1223_5

Vancouver

Sonkesriya A. Deep Learning-Based Signal Modulation Classification for Cognitive Radio Networks Using CNN-BiLSTM. International Journal of Recent Development in Engineering and Technology. 2023;12(12):20-28. doi:10.54380/ijrdet1223_5

BibTeX
@article{Sonkesriya2023,
  author = {Sonkesriya, Ajay},
  title = {Deep Learning-Based Signal Modulation Classification for Cognitive Radio Networks Using CNN-BiLSTM},
  journal = {International Journal of Recent Development in Engineering and Technology},
  year = {2023},
  volume = {12},
  number = {12},
  pages = {20--28},
  doi = {10.54380/ijrdet1223_5}
}
Kostenlos starten →