EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 2.5

Speech Enhancement by Empirical Mode Decomposition and Adaptive Noise Estimation

Kais Khalidi, Anis Mohamed

Defence Science Journal · 22.06.2026 · 76 (4) · 572-580

DOI: 10.14429/dsj.21670 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Defence Science Journal, online erschienen am 22.06.2026. Verfasst von Kais Khalidi und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 2.5).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This work introduces a fully adaptive speech denoising framework grounded in Empirical Mode Decomposition (EMD), designed to operate directly on the intrinsic time-frequency structure of the signal. The noisy speech waveform is adaptively separated into a set of intrinsic oscillatory modes, known as Intrinsic Mode Functions (IMFs), through an iterative time-domain procedure referred to as the sifting process. The idea of the proposed approach repose to threshold the IMFs adaptively. The main innovation of this work lies in the introduction of a time-adaptive IMF thresholding strategy driven by frame-level a posteriori SNR estimation. Adaptive noise is calculated directly from the noisy signal without the need for speech quiet detection. The enhanced speech waveform is obtained by recombining all adaptively thresholded IMFs, ensuring that each mode contributes only its cleaned oscillatory content. Because it relies solely on the intrinsic decomposition provided by Empirical Mode Decomposition, the proposed denoising framework operates as a fully datadriven and modelfree method. The proposed method is evaluated across multiple input SNR conditions and benchmarked against four classical baselines: EMDShrinkage, which applies a universal threshold, EMD-Kaiser, Improved EMD and adaptive wavelet thresholding (WTthresholding). The effectiveness of the proposed enhancement framework is assessed using both the output SignaltoNoise Ratio (SNRout) and the Perceptual Evaluation of Speech Quality (PESQ) metricExtensive simulations show that the proposed method provides stronger noise suppression and consistently outperforms other denoising methods, yielding higher SNR improvements and noticeably better perceptual quality. The present study focuses exclusively on speech signals degraded by additive white Gaussian noise.

Zitieren

APA

Khalidi, K., & Mohamed, A. (2026). Speech Enhancement by Empirical Mode Decomposition and Adaptive Noise Estimation. Defence Science Journal, 76(4), 572-580. https://doi.org/10.14429/dsj.21670

Vancouver

Khalidi K, Mohamed A. Speech Enhancement by Empirical Mode Decomposition and Adaptive Noise Estimation. Defence Science Journal. 2026;76(4):572-580. doi:10.14429/dsj.21670

BibTeX
@article{Khalidi2026,
  author = {Khalidi, Kais and Mohamed, Anis},
  title = {Speech Enhancement by Empirical Mode Decomposition and Adaptive Noise Estimation},
  journal = {Defence Science Journal},
  year = {2026},
  volume = {76},
  number = {4},
  pages = {572--580},
  doi = {10.14429/dsj.21670}
}
Kostenlos starten →