DocumentCode
2762451
Title
Speaker verification using a new representation based on a combination of MFCC and formants
Author
Mezghani, Ahmed ; O´Shaughnessy, Douglas
Author_Institution
Laboratoire d´´Electronique et des Technol. de l´´Inf.
fYear
2005
fDate
1-4 May 2005
Firstpage
1461
Lastpage
1464
Abstract
The purpose of this study is to investigate a new representation for automatic speaker verification using MFCC, formants and a combination of MFCC/formants. Verification is carried out using a weighted Euclidean distance also known as a Mahalanobis distance. The weighting coefficients are determined depending on their utility in the verification process. We report on extensive experiments that show the pertinence of the hybrid representation and the superior performance of the scheme
Keywords
fast Fourier transforms; speaker recognition; Mahalanobis distance; Mel scaled FFT based cepstrum; automatic speaker verification; hybrid representation; weighted Euclidean distance; Automatic speech recognition; Banking; Cepstral analysis; Cepstrum; Euclidean distance; Feature extraction; Hidden Markov models; Mel frequency cepstral coefficient; Speaker recognition; Telephony;
fLanguage
English
Publisher
ieee
Conference_Titel
Electrical and Computer Engineering, 2005. Canadian Conference on
Conference_Location
Saskatoon, Sask.
ISSN
0840-7789
Print_ISBN
0-7803-8885-2
Type
conf
DOI
10.1109/CCECE.2005.1557255
Filename
1557255
Link To Document