Title: Feature extraction and modelling techniques for multilingual speaker recognition: a review

Authors: B.G. Nagaraja; H.S. Jayanna

Addresses: Department of Electronics and Communication Engineering, GM Institute of Technology, Davangere 577006, Karnataka, India ' Department of Information Science and Engineering, Siddaganga Institute of Technology, Tumkur 572103, Karnataka, India

Abstract: In this paper, recognising the speaker in the multilingual context is demonstrated. Multilingual speaker recognition is useful in countries like India, Canada and South Africa where multiple languages are used for communication. In such countries, use of the monolingual speaker recognition system may not yield the expected level of performance. This is because a speaker may change his training and test languages whenever it is necessary. As a result, the system yields poor performance due to variation in language parameters. Therefore, developing a robust multilingual speaker recognition system is an issue in many countries. Researchers in the field of speaker recognition have made a few attempts to identify the speaker in a multilingual context. This paper gives an overview of various techniques developed for monolingual, cross-lingual and multilingual speaker recognition systems.

Keywords: speaker recognition; monolingual speakers; cross-lingual speakers; multilingual speakers; feature extraction; modelling; multiple languages.

DOI: 10.1504/IJSISE.2016.075000

International Journal of Signal and Imaging Systems Engineering, 2016 Vol.9 No.2, pp.67 - 78

Received: 25 Mar 2013
Accepted: 12 Mar 2014
Published online: 29 Feb 2016 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article