Two-stage signal denoising framework by singular spectrum analysis with Savitzky-Golay filter
Abstract
Context—Most signals encountered in real-world applications exhibit either nonlinear dynamics, or nonstationary behavior and they are frequently corrupted by several types of interference commonly termed as noise. In many practical applications, the direct utilization of raw noisy signals without appropriate enhancement or preprocessing can lead to substantial performance degradation in numerous signal processing applications such as feature extraction, pattern recognition, system identification, fault diagnosis, and parameter estimation. Therefore, effective signal processing techniques are required to mitigate noise and enhance considered signal quality.
Objective—Recovering an original signal from its contaminated version without any prior knowledge remains a fundamental and challenging problem in signal processing research. The presence of additive white Gaussian noise (AWGN), which commonly arises during signal acquisition, transmission, and storage processes degrades signal quality and obscures important underlying information. Consequently, the development of effective and user-independent noise reduction methods by researchers is inevitable. Thus, we consider the problem of denoising signals corrupted by AWGN using the singular spectrum analysis (SSA) with Savitzky-Golay (S-G) filter in this study.
Method—In the proposed method, noisy signal is decomposed into elementary signals whose energy is in decreasing order by SSA. To identify noise-dominant elementary signals in the distorted signal an adaptive threshold is defined as average of the singular values for the noise-dominant components. Partial recovery of the signal is executed residual noise in the recovered signal is smoothed by S-G filter. Performance of the proposed method is evaluated with three performance metrics over the five signals which are distorted by AWGN.
Results—Simulation results indicate that the proposed method achieves satisfactory denoising performance at particularly positive SNR levels since the SSA outputs are based on the energy distribution. When the input SNR is below 5 dB, the proposed approach provides an average SNR gain of about 20 dB. In addition, standard deviation results prove that the obtained results are reliable for single realization.
Conclusion—This study investigates the signal denoising problem using SSA framework. The proposed method requires neither a preprocessing nor a postprocessing stage. Experimental results on synthetic and real signals indicate that the proposed method is capable of effectively suppressing noise while preserving signal information under heavy noise conditions where the noise estimation is not possible. Future research may focus on developing an adaptive mechanism for determining the SSA frame length.
Keywords
References
- T. Qian, L. Zhang, Z. Li, “Algorithm of adaptive Fourier decomposition”, IEEE Transactions on Signal Processing, 59(12), 5899-5906, 2011. https://doi.org/10.1109/TSP.2011.2168520.
- T. Qian, “Adaptive Fourier decompositions and rational approximations—Part I: Theory”, International Journal of Wavelets, Multiresolution and Information Processing, 12(5), 1461008, 2014. https://doi.org/10.1142/S0219691314610086.
- Z. Wang, C. M. Wong, J. N. da Cruz, F. Wan, P.-I. Mak, P. U. Mak, “Muscle and electrode motion artifacts reduction in ECG using adaptive Fourier decomposition”, IEEE International Conference on Systems, Man and Cybernetics, San Diego, USA, 05-08 October 2014. https://doi.org/10.1109/SMC.2014.6974120.
- Z. Wang, F. Wan, C. M. Wong, L. Zhang, “Adaptive Fourier decomposition based ECG denoising”, Computers in Biology and Medicine, 77, 195-205, 2016. https://doi.org/10.1016/j.compbiomed.2016.08.013.
- A. Kizilkaya, A. Kirkbas, E. Bogar, “Signal denoising based on adaptive fourier decomposition”, 2017 Signal Processing: Algorithms, Architectures, Arrangements, and Applications, Poznan, Poland, 20-22 September 2017, 125-130. https://doi.org/10.23919/SPA.2017.8166851.
- N. E. Huang, Z. Shen, S. R. Long, M. L. C. Wu, H. H. Shih, Q. N. Zheng, N. C. Yen, C. C. Tung, H. H. Liu, “The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis”, Proceedings of the Royal Society A-Mathematical Physical and Engineering Sciences, 454(1971), 903-95, 1998. https://doi.org/10.1098/rspa.1998.0193.
- A.-O. Boudraa, J.-C. Cexus, “EMD-based signal filtering”, IEEE Transactions on Instrumentation and Measurement, 56(6), 2196-2202, 2007. https://doi.org/10.1109/TIM.2007.907967.
- T. -Y. Sun, C. -C. Liu, J. -H. Jheng, T.-Y. Tsai, “An efficient noise reduction algorithm using empirical mode decomposition and correlation measurement”, IEEE International Symposium on Intelligent Signal Processing and Communication System, Bangkok, Thailand, 08-11 February 2009, 286-289. https://doi.org/10.1109/ISPACS.2009.4806683.
Details
Primary Language
English
Subjects
Signal Processing
Journal Section
Research Article
Authors
Ali Kırkbaş
*
0000-0002-6402-8470
Türkiye
Early Pub Date
November 4, 2026
Publication Date
-
Submission Date
August 17, 2026
Acceptance Date
October 2, 2026
Published in Issue
Year 2026 Number: Advanced Online Publication