Spectrum Sensing and Waveform Classification with Deep Learning

Abstract

Over the years, engineers have gradually moved wireless communication systems upward into higher frequency bands, a trend driven by the need for wider bandwidth and larger coverage areas; radar, too, has long operated at these higher frequencies. Therefore, the spectrum of these two systems may overlap. In this research paper we designed a smart spectrum sensing system for radar and wireless communication system using deep convolutional neural network. We designed neural architecture for this purpose and labeling the data. Signal classification is equally important, because the growing number of modulation techniques makes it hard to tell at a glance whether a given burst is a wireless communication signal(smartphone call) or a whether radar ping. To tackle that challenge, the same neural architecture was adapted to recognize the waveform and modulation format of incoming signals, again relying on labeled synthetic data. MATLAB Simulink provided an accessible framework for generating, labeling, and pre-processing the training sets before they were fed into the network. For the spectrum-sensing task, traditional radar and wireless signals were synthesized in blocks, captured as time-series records, and then passed through the joint filter-classifier network. For waveform classification wigner ville distribution and convolutional neural network is used. Finally to the demonstrate and figure out the classification performance we used a confusion matrix, a confusion matrix compare true class values with predicted class values and gives the accurate classification values diagonally.

Introduction

Overview:

This study presents a deep learning-based approach to:

• Sense unused spectrum bands (spectrum sensing)

• Classify different radio waveforms and modulation types

The system processes signals from radar and commercial wireless communication systems (e.g., LTE, 5G), using deep convolutional neural networks (CNNs) within a MATLAB Simulink environment.

Research Objectives:

1. Identify and differentiate radar and wireless communication signals.

2. Classify waveform and modulation types using deep learning.

3. Improve detection of primary users (PUs) in specific frequency bands.

4. Ensure robustness against noise/interference.

5. Minimize misclassification and missed detections.

Literature Insights:

• Prior studies used machine learning (e.g., SVMs, Gaussian models) for spectrum sensing but faced limitations (e.g., noise uncertainty, complex implementation).

• Deep learning overcomes these by learning high-level features from large datasets without manual feature engineering.

• Application of DL in wireless communication is still emerging.

Methodology:

A. Spectrum Sensing:

• Environment: Simulated airport radar at 2.8 GHz and nearby LTE/5G signals with 30 scatterers.

• Data Preparation:

  • Used spectrogram images labeled as Radar, LTE, 5G, and Noise.

  • Split data: 80% training, 20% validation.

• Network Architecture:

  • Used ResNet50 as base for deeplabv3+ architecture.

  • Trained with SGDM algorithm.

• Testing & Evaluation:

  • Used semanticseg and confusion matrix to assess performance.

  • Achieved effective separation and identification of signal types.

B. Waveform Classification:

1. Radar Waveforms:

• Generated 3 modulation types:

  • Rectangular Pulse

  • Linear Frequency Modulation (LFM)

  • Barker Code

• Used Wigner-Ville Distribution (WVD) to extract time-frequency features.

• Stored features as 227×227 RGB images.

2. Wireless Waveforms:

• Generated 5 modulation types:

  • G-FSK, CP-FSK, B-FM, SSB-AM, DSB-AM

• WVD used for feature extraction.

• Data split: training, validation, testing.

3. Network Setup and Training:

• Used SqueezeNet for image classification.

• Converted WVD images for input.

• Trained CNN using trainNetwork in MATLAB.

• Training achieved ~96% accuracy, with 1280 iterations over 144 minutes.

Results:

• Spectrum Sensing: Accurate identification of radar, LTE, and 5G signals with clear signal labeling.

• Waveform Classification: High accuracy in detecting modulation types of both radar and communication signals.

• Confusion matrices confirm reliable classification and minimal misclassification.

Conclusion

Spectrum sensing and waveform classification are essential elements of contemporary wireless communication systems, facilitating effective spectrum use and reliable signal recognition. Utilizing deep learning methods, we have shown considerable progress in these fields. Deep learning architectures, like convolutional neural networks (CNNs), have demonstrated exceptional abilities in identifying intricate patterns within raw signal data, surpassing conventional techniques in accuracy and flexibility. Incorporating deep learning into spectrum sensing enhances the accuracy of detecting spectrum gaps and primary user signals, even in environments with low signal-to-noise ratio (SNR). Likewise, in waveform classification, deep learning models are proficient at recognizing and differentiating various modulation schemes.

References

[1] Abdalaziz Mohammad, FaroqAwinb, and Esam Abdel-Raheem. “Case study of TV spectrum sensing model based on machine learning techniques”, Ain Shams Engineering Journal, Volume 13, Issue 2, March 2022, 101540 [2] Awin FA, Alginahi YM, Abdel-Raheem E, Tepe K. Technical issues on cognitive radio-based internet of things systems: A survey.;7:97887–908, IEEE Access 2019 [3] Brynolfsson, Johan, and Maria Sandsten. \"Classification of one-dimensional non-stationary signals using the Wigner-Ville distribution in convolutional neural networks.\" 25th European Signal Processing Conference (EUSIPCO). IEEE, 2017. [4] Deep Learning in Mobile and Wireless Networking: A SurveyChaoyun Zhang, P. Patras, H. Haddadi, IEEE Communications Surveys & Tutorials 12 March 2018 [5] M. Zorzi, A. Zanella, A. Testolin, M. D. F. D. Grazia, and M. Zorzi, “Cognition-based networks: A new perspective on network optimization using learning and distributed intelligence,” IEEE Access, vol. 3, pp. 1512–1530, 2015. [6] Q. Cheng, Z. Shi, D. N. Nguyen, and E. Dutkiewicz, “Deep learning network based spectrum sensing methods for OFDM systems,” arXiv: 1807.09414, 2019. [7] G. Hinton et al., ‘‘Deep neural networks for acoustic modeling in speech recognition: The shared views of four research groups,’ IEEE Signal Process. Mag., vol. 29, no. 6, pp. 82–97, Nov. 2012 [8] Y. LeCun, Y. Bengio, and G. Hinton, ‘‘Deep learning,’’ Nature, vol. 521, pp. 436–44, May 2015. [9] R. Socher, Y. Bengio, and C. D. Manning. (2012). Deep Learning for NLP (Without Magic). [10] S. Zheng, P. Qi, S. Chen, and X. Yang, “Fusion methods for CNN-based automatic modulation classification,” IEEE Access, vol. 7, pp. 66496-66504, 2019. [11] C. Tanner Fredieu, Anthony Martone, R. Michael Buehrer, “Open-set Classification of Common Waveforms Using A Deep Feed-forward Network and Binary Isolation Forest Models”, Department of Electrical and Computer Engineering, Virginia Tech, Blacksburg, VA, USA,US Army Research Laboratory, Adelphi, MD, USA arXiv:2110.00252v1 [cs.LG] 1 Oct 2021 [12] F. F. Dighan, M.-S. Alouini, and M. K. Simon, “On the energy detection of unknown signals over fading channels,” in Proc. IEEE ICC, 2003, vol. 5, pp. 3575-3579. [13] Y. Zeng and Y.-C. Liang, “Eigenvalue-based spectrum sensing algorithms for cognitive radio,” IEEE Transactions on Communications, vol. 57, no. 6, pp. 1784-1793, 2009. [14] P.-H. Qi, Z. Li, J.-B. Si, et al., “A robust power spectrum split cancellation-based spectrum sensing method for cognitive radio systems,” Chinese Physics B, vol. 23, no. 12, pp. 128401 1-11, 2014. [15] K. M. Thilina, K. W. Choi, N. Saquib, and E. Hossain, “Machine learning techniques for cooperative spectrum sensing in cognitive radio networks,” IEEE Journal on Selected Areas in Communications, vol. 31, pp. 2209–2221, November 2013. [16] Kavita Bani and Vaishali Kulkarni, “Hybrid Spectrum Sensing Using MD and ED for Cognitive Radio Networks” Journal of Sensor and Actuator Networks 11(3):36 july, 2022 [17] R. Mei and Z. Wang, “Deep learning-based wideband spectrum sensing: A low computational complexity approach,” IEEE Communications Letters, vol. 27, no. 10, pp. 2633–2637, 2023. [18] W, Zhang, Y. Wang, X. Chen, Z. Cai, and Z. Tian, “Spectrum transformer: An attention-based wideband spectrum detector,” IEEE Transactions on Wireless Communications, vol. 23, no. 9, pp. 12?343–12?353, 2024 [19] S, Zheng, Z. Yang, W. Shen, L. Zhang, J. Zhu, Z. Zhao, and X. Yang, “Deep learning-based doa estimation,” IEEE Transactions on Cognitive Communications and Networking, vol. 10, no. 3, pp. 819–835, 2024. [20] Shilian Zheng, Zhihao Ye, Luxin Zhang, Keqiang Yue, and Zhijin Zhao “Deep Learning-Based Wideband Spectrum Sensing with Dual-Representation Inputs and Subband Shuffling Augmentation” License: CC BY-NC-SA 4.0 arXiv:2504.07427v1 [eess.SP] 10 Apr 2025 [21] E.V. Vijay, K. Aparna”RNN-BIRNN-LSTM based spectrum sensing for proficient data transmission in cognitive radio”. e-prime-advances in electrical engineering, Electronics and Energy, 6 (2023), Article 100378. [22] Syed, S.N.; Lazaridis, P.I.; Khan, F.A.; Ahmed, Q.Z.; Hafeez, M.; Holmes, V.; Chochliouros, I.P.; Zaharis, Z.D. Deep Learning Approaches for Spectrum Sensing in Cognitive Radio Networks. In Proceedings of the 2022 25th International Symposium on Wireless Personal Multimedia Communications (WPMC), Herning, Denmark, 30 October–2 November 2022; pp. 480–485. [23] Nallagonda, S.; Godugu, K.K.; Ranjeeth, M. Energy-Efficiency Analysis of Cognitive Radio Network with Improved Energy Detectors and SC Diversity over Nakagami-q Fading Environment. In Proceedings of the 2020 IEEE International Symposium on Sustainable Energy, Signal Processing and Cyber Security (iSSSC), Gunupur Odisha, India, 16–17 December 2020; pp. 1–6. [24] Ahmed, Q.Z.; Hafeez, M.; Khan, F.A.; Lazaridis, P. Towards Beyond 5G Future Wireless Networks with focus towards Indoor Localization. In Proceedings of the 2020 IEEE Eighth International Conference on Communications and Networking (ComNet), Hammamet, Tunisia, 27–30 October 2020; pp. 1–5. [25] Janu, D.; Singh, K.; Kumar, S. Machine learning for cooperative spectrum sensing and sharing: A survey. Trans. Emerg. Telecommun. Technol. 2021, 33, e4352. [26] Bkassiny, M.; Li, Y.; Jayaweera, S.K. A survey on machine-learning techniques in cognitive radios. IEEE Commun. Surv. Tutor. 2012, 15, 1136–1159. [27] Sherbin, K.; Sindhu, V. Cyclostationary Feature Detection for Spectrum Sensing in Cognitive Radio Network. In Proceedings of the 2019 International Conference on Intelligent Computing and Control Systems (ICCS), Madurai, India, 15–17 May 2019; IEEE: Piscataway Township, NJ, USA, 2019. [28] Vladeanu, C.; Martian, A.; Popescu, D.C. Spectrum Sensing with Energy Detection in Multiple Alternating Time Slots. IEEE Access 2022, 10, 38565–38574.

Copyright

Copyright © 2025 Muhammad Khan, Syed Muslim Shah. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.