Investigation on Inter-Speaker Variability in The Feature Space
description
Transcript of Investigation on Inter-Speaker Variability in The Feature Space
![Page 1: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/1.jpg)
Investigation on Inter-Speaker Variability in The Feature Space
Presenter : 陳彥達
![Page 2: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/2.jpg)
Reference
R. Haeb-Umbach, “Investigation on Inter-Speaker Variability in The Feature Space”, ICASSP 99.
![Page 3: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/3.jpg)
Outline Introduction A measure of inter-speaker variability Vocal tract normalization Cepstral mean and variance normalization
![Page 4: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/4.jpg)
Introduction Adaptation
Reduce mismatch by adapting feature vectors or model parameters to the target environment.
![Page 5: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/5.jpg)
Introduction(2) Normalization
Compute feature or model parameters that are insensitive to undesired variations of the speech signal.
![Page 6: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/6.jpg)
Introduction(3) Fisher discriminant analysis
An early assessment of a feature set without running recognition first
The ratio of feature variability due to different phonemes and due to different speakers
![Page 7: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/7.jpg)
A measure of inter-speaker variability
Good feature vector space Close together when belonging to the same
phoneme class Separated from each other when belonging to
the different phoneme class
![Page 8: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/8.jpg)
A measure of inter-speaker variability(2)
: cepstral feature vectors
: cepstral mean feature vector
: class mean vector
: total mean vector
)(tx
cphrtxcr
cr txN .,)(,
, )(1
CcN
mcphrrcr
cc ,,1;
1
}.|{,
C
c cphrr
C
ccccr mN
NNm
1 }.|{ 1,
11
![Page 9: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/9.jpg)
A measure of inter-speaker variability(3)
: cepstral mean feature vector : class mean vector
: total mean vector
: between class covariance matrix
: within class covariance matrix
m
cr ,
cm
Tccr
C
c cphrrccrW mm
NS )()(
1,
1 }.|{,
Tc
C
cc
cB mmmm
N
NS )()(
1
![Page 10: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/10.jpg)
A measure of inter-speaker variability(4)
Fisher variate analysis = the sum of the eigenvalues
of The radius of the scattering volume Higher
lower recognition error rate
)( 1BSStr
W
BW SS 1
)( 1BSStr
W
![Page 11: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/11.jpg)
Vocal tract normalization Reduce inter-speaker variability by a speaker-
specific frequency warping Differences in vocal tract length are compensated
for by a linear warping factor
)110(7001
)( 2595 melfk
melHz fkf
![Page 12: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/12.jpg)
Vocal tract normalization(2)
42 male + 42 female 42 male
![Page 13: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/13.jpg)
Vocal tract normalization(3)
a normalization on a per sentence basis performs better than a normalization on a per speaker basis
![Page 14: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/14.jpg)
Cepstral mean and variance normalization
: input cepstral feature : estimate of the mean of the input cepstral
feature : estimate of the standard deviation of the
input cepstral feature : the mean and variance normalized feature : number of features
Ddt
txtxty
d
ddd ,,1;
)(ˆ)()(
)(
)(tyd
)(txd)(txd
)(ˆ td
D
![Page 15: Investigation on Inter-Speaker Variability in The Feature Space](https://reader034.fdocuments.net/reader034/viewer/2022051622/5681557d550346895dc344e1/html5/thumbnails/15.jpg)
Cepstral mean and variance normalization(2)
42 male + 42 female 42 male