Feature extraction and normalization in SVM speaker verification using telephone speech

Master Thesis

2007

Permanent link to this Item
http://hdl.handle.net/11427/5167
Files
thesis_ebe_2007_mazibuko_t_t (1).pdf (4.9 MB)
Authors
Mazibuko, Thembisile Thulisile
Journal Title
Link to Journal
Journal ISSN
Volume Title
Publisher
Publisher

University of Cape Town

Department
Department of Electrical Engineering
Faculty
Faculty of Engineering and the Built Environment
License
Series
Abstract
In this research the Support Vector Machine classifier is applied to a text independent speaker verification task using conversational telephone speech from the NIST 2000 Speaker Recognition Evaluation. The SVM is a discriminative classifier with good generalization characteristics. It has been shown to perform as well as, and sometimes outperform the more widely used Gaussian Mixture Model. The SVM, like other classifiers is vulnerable to environmental noise, distortions from transmission over communication channels such as the telephone channel, and intersession variability.
Description

Includes bibliographical references (leaves 105-116).

Keywords
Electrical Engineering

Reference:

Mazibuko, T. 2007. Feature extraction and normalization in SVM speaker verification using telephone speech. University of Cape Town.

Collections
Masters