Multiple objective optimization applied to speech enhancement problem

Enhancement of speech corrupted by broadband noise is subject of interest in many applications to improve the intelligibility of speech and minimise the distortion. This paper describe a new effective method based on spectral noise subtracting technique. In fact, when using the latter technique, one...

Full description

Saved in:
Bibliographic Details
Published inICMIT : proceedings of the 2017 International Conference on Mathematics and Information Technology, Adrar, Algiers, December 4-5, 2017 pp. 24 - 28
Main Authors Ouznadji, Said, Chaabane, Djamal, Thameri, Messaoud
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.12.2017
Subjects
Distortion
multi-objective problem
Music
musical noise
Noise reduction
Optimization
particle swarm optimization
Random process
signal distortion
spectral subtraction
Speech
Speech enhancement
Speech recognition
speech signal enhancement
Online AccessGet full text
DOI10.1109/MATHIT.2017.8259691

Cover

More Information
Summary:Enhancement of speech corrupted by broadband noise is subject of interest in many applications to improve the intelligibility of speech and minimise the distortion. This paper describe a new effective method based on spectral noise subtracting technique. In fact, when using the latter technique, one introduce a new kind of noise referred to us musical noise. To overcome this situation, we extend to work proposed by Berouti et al. where the parameters of its technique (which are obtained empirically) will be obtained by optimizing a new function that realises a compromise between the enhancement of intelligibility of speech and the reduction of musical noise.
DOI:10.1109/MATHIT.2017.8259691