Deep learning based SNR estimation: Case study

Öz

Signal-to-noise ratio (SNR) estimation plays a critical role in optimizing wireless communication systems by enabling adaptive modulation, efficient power allocation, and reliable link adaptation. Traditional SNR estimation methods, whether data-aided or data-assisted, face significant challenges in sixth generation (6G) systems due to their high frequency, wide bandwidth, and heightened sensitivity to noise. In response to these challenges, deep learning (DL) models have emerged as a promising alternative. This study evaluates the SNR classification performance of three DL models—ResNet101V2, MobileNetV2, and Xception—utilizing transfer learning with star diagram images representing modulation types. The results indicate that ResNet101V2 achieves the highest classification accuracy of 70.8%, demonstrating robustness in handling high-order modulation types. MobileNetV2 achieves an accuracy of 63.6%, offering a viable alternative for resource-constrained scenarios due to its computational efficiency. In contrast, Xception, despite its established success in image classification tasks such as those on the ImageNet dataset, performs poorly in SNR classification with an accuracy of 56.8%. This disparity underscores the specificity of SNR classification as a unique challenge distinct from general image classification tasks. Additionally, as expected, classification accuracy declines with increasing modulation order, reflecting the complexity of higher-order modulations.

Anahtar Kelimeler

• SNR estimation
• Wireless communication
• Deep learning
• Transfer learning

Derin öğrenme tabanlı SNR kestirimi: Durum çalışması

Öz

Sinyal-gürültü oranı (signal to noise ratio, SNR) kestirimi, uyarlanabilir modülasyonu, etkili güç tahsisini ve güvenilir bağlantı uyarlamasını iyileştirdiği için kablosuz haberleşme sistemlerinin optimize edilmesinde önemli bir yere sahiptir. Veri yardımlı ve veri yardımsız olarak yapılan geleneksel SNR kestirim yöntemlerinin, yüksek frekans, geniş bant aralığı ve gürültüye karşı duyarlılığın fazla olması şeklinde karakterize edilen altıncı nesil (sixth generation, 6G) sistemlerinde yaşadıkları zorlukların aksine derin öğrenme (deep learning, DL) modelleri umut vaat eden bir alternatif olarak karşımıza çıkmaktadır. Bu çalışmada ResNet101V2, MobileNetV2 ve Xception olmak üzere üç adet DL modelinin SNR sınıflandırma performansı, modülasyon türlerine ait yıldız diyagramı görüntüleri yardımıyla öğrenme aktarımı tekniği kullanılarak değerlendirilmiştir. ResNet101V2, %70.8’lik bir ortalama sınıflandırma doğruluyla en üstün performansı gösterirken MobileNetV2 ve Xception sırasıyla %63.6 ve %56.8’lik doğruluk değerlerine ulaşabilmektedir. ResNet101V2, yüksek dereceli modülasyon türleri kullanılarak yapılan SNR sınıflandırmasında daha dayanıklı bir mimari olduğunu göstermiş olsa da MobileNetV2, kaynakları sınırlı senaryolar için alternatif olabilecek bir işlemsel yüke sahiptir. Tüm bunların aksine Xception, ImageNet veri setindeki görüntü sınıflandırma başarısına rağmen bu çalışmaya özgü olan SNR sınıflandırmasında aynı performansı gösterememektedir. Sonuçlar beklendiği üzere artan modülasyon derecesiyle beraber sınıflandırma doğruluğunun düştüğünü göstermektedir.

Anahtar Kelimeler

• SNR kestirimi
• Kablosuz haberleşme
• Derin öğrenme
• Öğrenme aktarımı

Kaynakça

1. Z. Zhou, A. Kassem, J. Seddon, E. Sillekens, I. Darwazeh, P. Bayvel, Z. Liu, 938 Gb/s, 5–150 GHz ultra-wideband transmission over the air using combined electronic and photonic-assisted signal generation, Journal of Lightwave Technology, 42(20), 7247–7252, 2024. https://doi.org/10.1109/JLT.2024. 3446827.
2. H. Abeida, T. Y. Al-Nafouri, S. Al-Ghadhban, Data-aided SNR estimation in time-variant Rayleigh fading channels. In 2010 IEEE 11th International Workshop on Signal Processing Advances in Wireless Communications (SPAWC), IEEE, pp. 1–5, 2010. https://doi.org/10.1109/SPAWC.2010.5671005.
3. R. Matzner, F. Englberger, An SNR estimation algorithm using fourth-order moments. In Proceedings of 1994 IEEE International Symposium on Information Theory, IEEE, pp. 119, 1994. https://doi.org/10.1109 /ISIT.1994.394869.
4. T. Xu, I. Darwazeh, Wavelet classification for non-cooperative non-orthogonal signal communications. In 2020 IEEE Globecom Workshops, GC Wkshps 2020-Proceedings, Institute of Electrical and Electronics Engineers Inc., pp. 1-6, 2020. https://doi.org/10.1109 /GCWkshps50303.2020.9367556.
5. Y. Liu, H. Du, D. Niyato, J. Kang, Z. Xiong, D.I. Kim, A. Jamalipour, Deep generative model and its applications in efficient wireless network management: A tutorial and case study, IEEE Wireless Communications, pp. 199-207, 2024. https://doi.org/ 10.1109/MWC.009.2300165.
6. H. Ye, L. Liang, G. Y. Li, B. H. Juang, Deep learning-based end-to-end wireless communication systems with conditional GANs as unknown channels, IEEE Transactions on Wireless Communications, 19(5), 3133–3143, 2020. https://doi.org/10.1109/TWC.2020 .2970707.
7. S. M. Aldossari, K. C. Chen, Machine learning for wireless communication channel modeling: An overview. Wireless Personal Communications, 106, 41–70, 2019. https://doi.org/10.1007/s11277-019-06275-4.
8. X. Xie, S. Peng, X. Yang, Deep learning‐based signal‐to‐noise ratio estimation using constellation diagrams, Mobile Information Systems 2020(1), 8840340, 2020. https://doi.org/10.1155/2020/8840340.

9. T. Ngo, B. Kelley, P. Rad, Deep learning based prediction of signal-to-noise ratio (SNR) for LTE and 5G systems, In 2020 8th International Conference on Wireless Networks and Mobile Communications (WINCOM), IEEE, pp. 1–6, 2020. https://doi.org /10.1109/WINCOM50532.2020.9272470.
10. D. Athanasios, G. Kalivas, SNR estimation for low bit rate OFDM systems in AWGN channel, In International Conference on Networking, International Conference on Systems and International Conference on Mobile Communications and Learning Technologies (ICNICONSMCL’06), IEEE, pp. 198-198, 2006. https://doi.org/10.1109/ICNICONSMCL. 2006.198.
11. D. Guo, Y. Wu, S. Shamai, S. Verdú, Estimation in Gaussian noise: Properties of the minimum mean-square error, IEEE Transactions on Information Theory 57(4), 2371–2385 2011. https://doi.org/10.1109/TIT. 2011.2111010.
12. S.K. Tiwari, P.K. Upadhyay, Maximum likelihood estimation of SNR for diffusion-based molecular communication, IEEE Wireless Communications Letters 5(3), 320-323, 2016. https://doi.org/10.1109 /LWC.2016.2553034.
13. R. Gagliardi, C. Thomas, PCM data reliability monitoring through estimation of signal-to-noise ratio, IEEE Transactions on Communications 16(3), 479–486, 1968. https://doi.org/10.1109/TCOM.1968.108 9851.
14. D. R. Pauluzzi, N.C. Beaulieu, A comparison of SNR estimation techniques for the AWGN channel, IEEE Transactions on Communications 48(10), 1681–1691, 2000. https://doi.org/10.1109/26.871393.
15. T. Salman, A. Badawy, T. M. Elfouly, T. Khattab, A. Mohamed, Non-data-aided SNR estimation for QPSK modulation in AWGN channel, In 2014 IEEE 10th International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob), IEEE, pp. 611-616, 2014. https://doi.org /10.1109/WiMOB.2014.6962233.
16. B. Shah, S. Hinedi, The split symbol moments SNR estimator in narrow-band channels, IEEE transactions on aerospace and electronic systems, 26(5), 737-747, 1990. https://doi.org/10.1109/7.102709.
17. M. K. Simon, A. Mileant, SNR estimation for the baseband assembly, The Telecommunications and Data Acquisition Report,1986.
18. A. L. Brandao, L. B. Lopes, D. C. McLemon, In-service monitoring of multipath delay and cochannel interference for indoor mobile communication systems, In Proceedings of ICC/SUPERCOMM’94 - 1994 International Conference on Communications, IEEE, pp. 1458–1462, 1994. https://doi.org/10.1109/ICC. 1994.368788.
19. K. Yang, Z. Huang, X. Wang, F. Wang, An SNR estimation technique based on deep learning, Electronics, 8(10), 1139, 2019. https://doi.org/10.3390 /electronics8101139.
20. H. Li, D.L. Wang, X. Zhang, G. Gao, Frame-level signal-to-noise ratio estimation using deep learning, In Proceedings of the Annual Conference of the International Speech Communication Association, INTERSPEECH, International Speech Communication Association, pp. 4626–4630, 2020. https://doi.org /10.21437/Interspeech.2020-2475.
21. S. Jeevangi, S. Jawaligi, V. Patil, Deep Learning-based SNR Estimation for Multistage Spectrum Sensing in Cognitive Radio Networks, Journal of Telecommunications and Information Technology, (4), 21–31, 2022. https://doi.org/10.26636/jtit.2022.1649 22.
22. S. Zheng, S. Chen, T. Chen, Z. Yang, Z. Zhao, X. Yang, Deep Learning-Based SNR Estimation, IEEE Open Journal of the Communications Society, (5), 4778-4796, 2024. https://doi.org/10.1109/OJCOMS.2024. 3436640.
23. S. Chen, S. Zheng, Z. Yang, T. Chen, Z. Zhao, X. Yang, Deep Learning-Based SNR Estimation with Covariance Input, In International Conference on Communication Technology Proceedings, ICCT, Institute of Electrical and Electronics Engineers Inc., pp. 181–187, 2023 https://doi.org/10.1109/ ICCT59356.2023.10419442.
24. B. Xu, T. Ding, L. Guo, AC-BiLSTM: A Spatial Bidirectional LSTM with Multi-Head Self-Attention for SNR Estimation, In 2024 4th International Conference on Computer Systems, ICCS 2024, Institute of Electrical and Electronics Engineers Inc., pp. 34–38, 2024. https://doi.org/10.1109/ICCS62594. 2024.10795825.
25. D. Hu, Y. Zhao, W.J. Xie, Q. Xiao, L. Li, A squeeze-and-excitation network for SNR estimation of communication signals, IET Communications, 19(1) 2025. https://doi.org/10.1049/cmu2.70006.