Araştırma Makalesi

A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification

Cilt: 14 28 Temmuz 2026
PDF İndir
EN TR

A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification

Öz

Signals produced by physical systems often exhibit complex time-varying structures that reflect the internal dynamics of engineering processes. Accurate interpretation of such signals is essential for diagnostics, monitoring, and datadriven decision making in applications ranging from biomedical monitoring to industrial systems and renewable energy platforms. Conventional spectral analysis methods provide valuable frequency-domain information but are often insufficient for analyzing transient and non-stationary signal behavior. This paper presents a unified mathematical framework for time–frequency feature learning in physical signal classification. The proposed formulation establishes a systematic connection between signal representations, feature construction mechanisms, and statistical learning modelsthe transient and non-stationary characteristics ofsformed into time–frequency representations using operators such as the Short-Time Fourier Transform and the Continuous Wavelet Transform. Informative signal descriptors are then derived through energy-based measures and entropy-based features that characterize the structural distribution of signal energy across the time–frequency plane. The resulting descriptors are organized into compact feature vectors that provide a structured representation suitable for machine learning algorithms. The framework formalizes the mapping between signal space, time–frequency representation space, and feature space, enabling consistent integration of signal processing techniques with classification models such as artificial neural networks. The proposed formulation provides a general analytical structure for extracting discriminative patterns from complex engineering signals and supports systematic development of signal-based classification systems across multiple application domains.

Anahtar Kelimeler

Kaynakça

  1. [1] H. Azami and A. Fernández, "Time-frequency feature extraction for machine learning-based signal classification: A review," Information Sciences, vol. 553, pp. 1-27, 2021, doi: 10.1016/j.ins.2020.12.048.
  2. [2] A. K. Singh and B. Singh, "Time-frequency analysis of non-stationary power system signals for improved monitoring and diagnostics," IEEE Transactions on Industrial Informatics, vol. 18, no. 9, pp. 6124-6134, 2022, doi: 10.1109/TII.2021.3139874.
  3. [3] S. S. R. Mohanty, P. K. Dash, and B. K. Panigrahi, "A comprehensive review on non-stationary signal processing techniques for condition monitoring in power systems," IEEE Systems Journal, vol. 15, no. 3, pp. 3950-3961, 2021, doi: 10.1109/JSYST.2020.3048104.
  4. [4] Y. Zhang, X. Li, and Z. Chen, "Transient feature extraction of engineering signals using adaptive time-frequency decomposition," IEEE Transactions on Instrumentation and Measurement, vol. 71, pp. 1-12, 2022, doi: 10.1109/TIM.2022.3156789.
  5. [5] M. A. A. Pedersen and J. M. Hansen, "Non-stationary signal analysis for structural health monitoring: A modern review," Mechanical Systems and Signal Processing, vol. 165, 108300, 2022, doi: 10.1016/j.ymssp.2021.108300.
  6. [6] X. Su, D. Shi, B. Wu, W.-S. Gan, L. Ye, and K. Song, "SNR enhancement for non-stationary signals in arrays via ring model and message passing," Mechanical Systems and Signal Processing, vol. 237, 113092, 2025, doi: 10.1016/j.ymssp.2025.113092.
  7. [7] H. Miao and J. Shi, "Generalized singular spectrum analysis associated with fractional Fourier transform," IEEE Transactions on Signal Processing, doi: 10.1109/TSP.2026.3667516.
  8. [8] A. Hemati, "Experimental investigation on bearing pitting fault diagnosis using multi-domain vibration analysis and advanced signal processing," Journal of Failure Analysis and Prevention, vol. 26, pp. 422-436, 2026, doi: 10.1007/s11668-025-02361-x.

Ayrıntılar

Birincil Dil

İngilizce

Konular

Elektrik Mühendisliği (Diğer)

Bölüm

Araştırma Makalesi

Yayımlanma Tarihi

28 Temmuz 2026

Gönderilme Tarihi

14 Mart 2026

Kabul Tarihi

31 Mart 2026

Yayımlandığı Sayı

Yıl 2026 Cilt: 14

Kaynak Göster

APA
Akgün, Ö. (2026). A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification. Balkan Journal of Electrical and Computer Engineering, 14. https://doi.org/10.17694/bajece.1909911
AMA
1.Akgün Ö. A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification. Balkan Journal of Electrical and Computer Engineering. 2026;14. doi:10.17694/bajece.1909911
Chicago
Akgün, Ömer. 2026. “A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification”. Balkan Journal of Electrical and Computer Engineering 14 (Temmuz). https://doi.org/10.17694/bajece.1909911.
EndNote
Akgün Ö (01 Temmuz 2026) A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification. Balkan Journal of Electrical and Computer Engineering 14
IEEE
[1]Ö. Akgün, “A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification”, Balkan Journal of Electrical and Computer Engineering, c. 14, Tem. 2026, doi: 10.17694/bajece.1909911.
ISNAD
Akgün, Ömer. “A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification”. Balkan Journal of Electrical and Computer Engineering 14 (01 Temmuz 2026). https://doi.org/10.17694/bajece.1909911.
JAMA
1.Akgün Ö. A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification. Balkan Journal of Electrical and Computer Engineering. 2026;14. doi:10.17694/bajece.1909911.
MLA
Akgün, Ömer. “A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification”. Balkan Journal of Electrical and Computer Engineering, c. 14, Temmuz 2026, doi:10.17694/bajece.1909911.
Vancouver
1.Ömer Akgün. A Unified Mathematical Framework for Time–Frequency Feature Learning in Physical Signal Classification. Balkan Journal of Electrical and Computer Engineering. 01 Temmuz 2026;14. doi:10.17694/bajece.1909911

All articles published by BAJECE are licensed under the Creative Commons Attribution 4.0 International License. This permits anyone to copy, redistribute, remix, transmit and adapt the work provided the original work and source is appropriately cited.Creative Commons Lisans