Analysis of Dense Neural Network architectures for signal classiɯcation based on complex samples
DOI:
https://doi.org/10.34169/2414-0651.2025.3(47).57-64Keywords:
signal classification, Dense Neural Network, power spectral density, Welch method, complex signal samples, amplitude, phase, statistical features, deep learningAbstract
This study presents an analysis of the effectiveness of the Dense Neural Network (DNN) deep learning architecture in the task of radio signal classification using features derived from complex baseband samples (in-phase and quadrature components). The input feature vector was formed using the power spectral density estimated via the Welch method, complemented by additional statistical, amplitude-phase and frequency-energy characteristics of the signal. The model was trained, validated and tested on a unified feature set to ensure objective evaluation. Performance was assessed using accuracy, recall, F1-score metrics and confusion matrices. Experimental results demonstrated that the Dense Neural Network can achieve high classification accuracy even under varying signal characteristics. The proposed approach can be applied in automated systems for radio frequency spectrum monitoring and analysis, as well as in tasks related to signal detection and identification in complex electromagnetic environments.
Downloads
References
Kumar, S., Ahmadb, I., Höyhtyäc, M., Khand, S. & Gurtove, A. Deep Learning Frameworks for Cognitive Radio Networks: Review and Open Research Challenges. Available at: https://arxiv.org/pdf/2410.23949.
Kim, S., Kim, J., Doan, V. & Kim, D. Lightweight Deep Learning Model for Automatic Modulation Classification in Cognitive Radio Networks. Available at: https://www.researchgate.net/publication/346433098_Lightweight_Deep_Learning_Model_for_Automatic_Modulation_Classification_in_Cognitive_Radio_Networks.
Fekry, O., Abdalla, M. & Elsayed, A. Deep Learning-Based Automatic Modulation Classification Using Robust CNN Architecture for Cognitive Radio Networks. Available at: https://www.researchgate.net/publication/346433098_Lightweight_Deep_Learning_Model_for_Automatic_Modulation_Classification_in_Cognitive_Radio_Networks.
O’Shea,T. J., Roy, T. & Clancy, T. C. Over-the-Air Deep Learning Based Radio Signal Classification. Available at: https://arxiv.org/abs/1712.04578.
O’Shea, J. Hoydis. An Introduction to Deep Learning for the Physical Layer. Available at: https://arxiv.org/abs/1702.00832.
Welch, P. The Use of Fast Fourier Transform for the Estimation of Power Spectra: A Method Based on Time Averaging Over Short, Modified Periodograms. Available at: https://ieeexplore.ieee.org/document/1161901.
Krizhevsky, H., Sutskever, I. & Hinton, G. E. ImageNet Classification with Deep Convolutional Neural Networks. Available at: https://dl.acm.org/doi/10.1145/3065386.
Chollet, F. Keras: Deep Learning Library for Python. Available at: https://keras.io.
M. Abadi et al. TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems. Available at: https://www.tensorflow.org.
Haykin, S. Communication Systems. Available at: https://onlinelibrary.wiley.com/doi/book/10.1002/9781119474744.
SciPy documentation – Signal Processing. Available at: https://docs.scipy.org/doc/scipy/reference/signal.html.
IQ Data Explained. PE0SAT Satellite Ground Station : Available at: https://www.pe0sat.vgnet.nl/sdr/iq-data-explained/.
IQ Signal Master Vector Signal Analysis Software. Anritsu : Available at: https://www.anritsu.com/en-us/test-measurement/products/mx280005a.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Вадим Слюсар,Вадим Козлов,Денис Козлов
This work is licensed under a Creative Commons Attribution 4.0 International License.
