A New Approach for Blind Speech Separation Based on Sinusoidal Model of Speech Signals
IEEE Transactions on Signal Processing · 2026 · 74 · 1631-1640
DOI: 10.1109/tsp.2026.3669182 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Signal Processing, 2026. Verfasst von Sadra Siahpoosh und 1 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 33 Literaturangaben.
Abstract (Originalsprache)
This paper introduces a new method for blind speech separation in both instantaneous and convolutive mixtures using dual-microphone setups. The proposed approach is also applicable to underdetermined mixtures, where the number of sources exceeds the number of microphones. The method combines sinusoidal speech model analysis and synthesis with spatial and diarization-driven cues to facilitate the separation of speech signals. By clustering sinusoids and synthesizing separated speech signals, this method shows potential for achieving effective speech separation. The approach is evaluated using stereo mixtures from SiSEC 2013 dataset. The experimental results indicate that the method excels in underdetermined instantaneous mixtures and performs comparably in convolutive mixture scenarios.
Zitieren
APA
Siahpoosh, S., & Babaie-Zadeh, M. (2026). A New Approach for Blind Speech Separation Based on Sinusoidal Model of Speech Signals. IEEE Transactions on Signal Processing, 74, 1631-1640. https://doi.org/10.1109/tsp.2026.3669182
Vancouver
Siahpoosh S, Babaie-Zadeh M. A New Approach for Blind Speech Separation Based on Sinusoidal Model of Speech Signals. IEEE Transactions on Signal Processing. 2026;74:1631-1640. doi:10.1109/tsp.2026.3669182
BibTeX
@article{Siahpoosh2026,
author = {Siahpoosh, Sadra and Babaie-Zadeh, Massoud},
title = {A New Approach for Blind Speech Separation Based on Sinusoidal Model of Speech Signals},
journal = {IEEE Transactions on Signal Processing},
year = {2026},
volume = {74},
pages = {1631--1640},
doi = {10.1109/tsp.2026.3669182}
}
EUVIMED