Automatic modulation classification using ResNet with signal graph and discrete wavelet transform-based features

Öz

In this study, a hybrid approach is proposed for automatic modulation classification by integrating signal representations with discrete wavelet transform-based features using the ResNet-50 deep learning architecture. The primary objective of the study is to enhance the classification performance of modulation schemes in noisy communication channels by jointly utilizing both time-domain and frequency-domain characteristics of the signals. Simulations were conducted on four fundamental modulation types—OOK, BPSK, QPSK, and 16QAM—selected from the RadioML 2018.01A dataset. For each signal, detail coefficients representing high-frequency components were extracted using the Coiflet-2 (coif2) wavelet. Subsequently, these coefficients were combined with the observed signal representations, and the resulting data were provided as input to the deep learning model. Simulation results obtained under various signal-to-noise ratio (SNR) conditions demonstrate that the proposed model operates robustly even at low SNR levels and achieves high classification accuracy. As expected, the accuracy improves with increasing SNR, reaching over 98% at 30 dB. These findings indicate that wavelet-based features significantly enhance the model’s ability to capture both structural and transient characteristics of signals, thereby reducing the confusion between similar modulation types.

Anahtar Kelimeler

• Discrete wavelet transform
• Deep learning
• Modulation classification
• ResNet-50
• Feature extraction

İşaret grafiği ve ayrık dalgacık dönüşümü tabanlı öznitelikler ile ResNet kullanarak otomatik modülasyon sınıflandırması

Öz

Bu çalışmada, otomatik modülasyon sınıflandırması için ResNet-50 derin öğrenme mimarisi ile işaret grafiği ve ayrık dalgacık dönüşümü tabanlı özniteliklerin birleştirilmesi yaklaşımı önerilmiştir. Çalışmanın amacı, gürültülü haberleşme kanallarında modülasyon türlerinin sınıflandırma başarımını artırmak amacıyla sinyallerin hem zaman hem de frekans alanı özelliklerini birlikte kullanmaktır. RadioML 2018.01A veri setinden seçilen dört temel modülasyon türü (OOK, BPSK, QPSK ve 16QAM) üzerinde benzetimler gerçekleştirilmiştir. Her bir sinyal için Coiflet-2 (coif2) dalgacığı kullanılarak yüksek frekans bileşenlerini temsil eden detay katsayıları elde edilmiştir. Daha sonra, bu katsayılar ile gözlenen işaretin grafikleri birleştirilerek derin öğrenme modeline girdi olarak verilmiştir. Farklı sinyal gürültü oranı koşullarında yapılan benzetimlerde, önerilen modelin düşük sinyal-gürültü oranı değerlerinde dahi kararlı bir şekilde çalıştığı ve yüksek doğruluk oranlarına ulaştığı gözlenmiştir. Sinyal-gürültü oranı arttıkça doğruluk oranında beklendiği üzere yükselme gerçekleşmiş ve 30 dB seviyesinde %98’in üzerinde bir sınıflandırma başarısı elde edilmiştir. Bu sonuçlar, dalgacık tabanlı özniteliklerin modelin hem yapısal hem de geçici sinyal özelliklerini yakalama kabiliyetini güçlendirdiğini ve benzer modülasyon türleri arasındaki karışıklığı azalttığını göstermektedir.

Anahtar Kelimeler

• Ayrık dalgacık dönüşümü
• Derin öğrenme
• Modülasyon sınıflandırması
• ResNet-50
• Öznitelik çıkarımı

Kaynakça

1. Zhang, H., Huang, M., Yang, J., and Sun, W. (2021) A data preprocessing method for automatic modulation classification based on CNN, IEEE Communications Letters, 25(4): 1206-1210.
2. Jiang, M., Quan, D., Zhou, F., Yu, K., Chen, Y., and Jin, N. (2025) Modulation recognition of radar signals based on multimodal contrastive learning, IEEE Journal of Microwaves, 5(5): 1082-1093.
3. Jafar, N., Paeiz, A., and Farzaneh, A. (2021) Automatic modulation classification using modulation fingerprint extraction, Journal of Systems Engineering and Electronics, 32(4): 799-810.
4. Dobre, O. (2015) Signal identification for emerging intelligent radios: classical problems and new challenges, IEEE Instrumentation & Measurement Magazine, 18(2): 11-18.
5. Kumar, A., Manish, and Satija, U. (2023) Residual stack-aided hybrid CNN-LSTM-based automatic modulation classification for orthogonal time-frequency space system, IEEE Communications Letters, 27(12): 3255-3259.
6. An, T.T. and Lee, B.M. (2023) Robust automatic modulation classification in low signal to noise ratio, IEEE Access, 11: 7860-7872.
7. Zheng, S., Qi, P., Chen, S., and Yang, X. (2019) Fusion methods for CNN-based automatic modulation classification, IEEE Access, 7: 66496-66504.
8. Gehlot, S., Kumar, N., Srivastava, S., and Yadav, S.K. (2024) Modulation classification in NOMA systems using lightweight dual-pooling CNN with superposed constellation density grids (SCDGs). 2024 IEEE 100th Vehicular Technology Conference (VTC2024-Fall), 1-5.

9. Kim, B., Sathyanarayanan, V., Mecklenbräuker, C., and Gerstoft, P. (2023) Deep learning-based modulation classification for OFDM systems without symbol-level synchronization. 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing Workshops (ICASSPW), 1-5.
10. Nagpal, P., Bhinge, S.A., and Shitole, A. (2022) A comparative analysis of ResNet architectures. 2022 International Conference on Smart Generation Computing, Communication and Networking (SMART GENCON), 1-8.
11. Ghosh, T. and Jayanthi, N. (2024) An efficient Dense-ResNet for multimodal image fusion using medical image, Multimedia Tools and Applications, 83(26): 68181-68208.
12. Arun M., Deenadhayalan M., Deepak R., Balaraman S. and Deepak D. (2024) Lung cancer detection using ResNet-50 CNN architecture. 2024 Second International Conference on Advances in Information Technology (ICAIT), 1-6.
13. Jaber, M.M., Ali, M.H., Abd, S.K., Jassim, M.M., Alkhayyat, A., Majid, M.Sh., Alkhuwaylidee, A.R., and Alyousif, S. (2023) Resnet-based deep learning multilayer fault detection model-based fault diagnosis, Multimedia Tools and Applications, 83(7): 19277-19300.
14. He, K., Zhang, X., Ren, S., and Sun, J. (2016) Deep residual learning for image recognition. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 770-778.
15. Liang, J. (2020) Image classification based on RESNET, Journal of Physics: Conference Series, 1634(1): 012110.
16. Hashmi, M.F., Devipriyanka, M., Raghavaram, B., and Husain, A.A. (2025) Advanced semantic lung segmentation with a hybrid SegNet-ResNet50 network, Journal of Electrical Systems and Information Technology, 12(1): 64.
17. Tripathy, S., Pattanaik, S.R., Rabindranath, M.H., and Dash, R. (2024) Enhanced textual image classification with ensemble learning of MobileNetV2 and ResNet50. 2024 International Conference on Intelligent Computing and Sustainable Innovations in Technology (IC-SIT), 1-5.
18. Dasari, S., Poonguzhali, B., and Rayudu, M. (2023) Transfer learning approach for classification of diabetic retinopathy using fine-tuned ResNet50 deep learning model. 2023 International Conference on Sustainable Communication Networks and Application (ICSCNA), 1361-1367.
19. Wei, D. and Bovik, A.C. (1998) Sampling approximation of smooth functions via generalized coiflets, IEEE Transactions on Signal Processing, 46(4): 1133-1138.
20. Narkhedkar, S.G. and Patel, P.K. (2014) Recipe of speech compression using coiflet wavelet. 2014 International Conference on Contemporary Computing and Informatics (IC3I), 1135-1139.
21. Rohima and Akbar, M.B. (2020) Wavelet analysis and comparison from Coiflet family on image compression. 2020 8th International Conference on Cyber and IT Service Management (CITSM), 1-5.
22. Joseph, S.M. and Babu, A.P. (2013) Continuous speech coding using coiflets wavelet. 2013 International Conference on Signal Processing, Image Processing & Pattern Recognition, 253-257.
23. Guo, T., Zhang, T., Lim, E., Lopez-Benitez, M., Ma, F., and Yu, L. (2022) A review of wavelet analysis and its applications: challenges and opportunities, IEEE Access, 10: 58869-58903.
24. DeepSig Inc. (2018) RadioML 2018.01A Dataset.
25. Ouamna, H., Lakouismi, C., Madini, Z., and Zouine, Y. (2024) Deep learning based automatic modulation classification using IQ signals. 2024 10th International Conference on Optimization and Applications (ICOA), 1-5.
26. O’Shea, T.J., Roy, T., and Clancy, T.C. (2018) Over-the-air deep learning based radio signal classification, IEEE Journal of Selected Topics in Signal Processing, 12(1): 168-179.
27. Dixit, A. and Majumdar, S. (2013) Comparative analysis of Coiflet and Daubechies wavelets using global threshold for image de-noising, International Journal of Advances in Engineering & Technology (IJAET), 6(5): 2247-2252.
28. Akbar, M.B. (2020) Comparison and analysis several wavelet from Daubechies family in image compression. 2020 8th International Conference on Cyber and IT Service Management (CITSM), 1-4.
29. Zhen, P., Zhang, B., Chen, Z., Guo, D., and Ma, W. (2022) Spectrum sensing method based on wavelet transform and residual network, IEEE Wireless Communications Letters, 11(12): 2517-2521.
30. Kaur, J. and Rani, S. (2025) Transforming pistachio classification: the power of ResNet-50 and SVM integration. 2025 International Conference on Ambient Intelligence in Health Care (ICAIHC), 1-6.