27 0

Similar speaker recognition using nonlinear analysis

Title
Similar speaker recognition using nonlinear analysis
Author
권영헌
Issue Date
2004-07
Publisher
PERGAMON-ELSEVIER SCIENCE LTD
Citation
CHAOS, SOLITONS & FRACTALS, v. 21, No. 1, Page. 159-164
Abstract
Speech features of the conventional speaker identification system, are usually obtained by linear methods in spectral space. However, these methods have the drawback that speakers with similar voices cannot be distinguished, because the characteristics of their voices are also similar in spectral space. To overcome the difficulty in linear methods, we propose to use the correlation exponent in the nonlinear space as a new feature vector for speaker identification among persons with similar voices. We show that our proposed method surprisingly reduces the error rate of speaker identification system to speakers with similar voices.
URI
https://www.sciencedirect.com/science/article/pii/S0960077903005459#!https://repository.hanyang.ac.kr/handle/20.500.11754/149362
ISSN
0960-0779
DOI
10.1016/j.chaos.2003.10.032
Appears in Collections:
COLLEGE OF SCIENCE AND CONVERGENCE TECHNOLOGY[E](과학기술융합대학) > APPLIED PHYSICS(응용물리학과) > Articles
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE