Single Channel Speech Enhancement Using Masking Based on Sinusoidal Modeling

This paper focused on development of single channel speech enhancement method. Conventional noise reduction methods based on filtering like Wiener filtering and masking uses spectral magnitudes. These magnitudes are obtained from time-frequency representation of noisy speech signals. Here, speech si...

Full description

Saved in:
Bibliographic Details
Published inRecent Trends in Image Processing and Pattern Recognition Vol. 1576; pp. 330 - 337
Main Authors Buragohain, Rantu, Reddy, R. Aditya, Venkatesh, Yenduri, Prabhakar, Gudmalwar Ashishkumar, Rao, Ch. V. Rama
Format Book Chapter
LanguageEnglish
Published Switzerland Springer International Publishing AG 2022
Springer International Publishing
SeriesCommunications in Computer and Information Science
Subjects
Online AccessGet full text
ISBN3031070046
9783031070044
ISSN1865-0929
1865-0937
DOI10.1007/978-3-031-07005-1_28

Cover

More Information
Summary:This paper focused on development of single channel speech enhancement method. Conventional noise reduction methods based on filtering like Wiener filtering and masking uses spectral magnitudes. These magnitudes are obtained from time-frequency representation of noisy speech signals. Here, speech signal is analyzed using sinusoidal modelling. Filter gain is developed for masking of the background noise based on sinusoidal components. The developed system’s performance is evaluated using Perceptual Evaluation of Speech Quality (PESQ). It is evident from experiments that proposed approach displaying better performance compared to existing approaches.
ISBN:3031070046
9783031070044
ISSN:1865-0929
1865-0937
DOI:10.1007/978-3-031-07005-1_28