مشاهده مشخصات مقاله
Efficient methods for decreasing training and testing times in speaker verification systems
|
Authors |
-
M. Raissi Dehkordi
-
M.M. Homayonpour
|
|
Conference |
دوازدهمین کنفرانس بینالمللی سالانه انجمن کامپیوتر ایران |
|
Abstract |
Gaussian Mixture Model (GMM) models feature space using mean vector and covariance matrix of probability
density functions of feature vectors. In this paper, improved Vector Quantization and Covariance Matrix methods were
compared to GMM. Covariance Matrix model considers shape of probability density functions, while Vector
quantization method models position of probability density functions in feature space. In this paper different methods of
Covariance matrix model including Arithmetic-harmonic sphericity measure and Divergence Shape measure were
examined to evaluate scores in speaker verification task.. Experimental results show that Arithmetic-harmonic
sphericity measure outperforms Divergence shape measure. A novel vector quantization approach was also presented in
this paper. This approach is based on comparing codebook obtained from training data to codebook obtained from test
data. Results show that recent approach has a better performance compared to traditional vector quantization approach.
Also, the results show that Covariance matrix model outperforms improved Vector Quantization and GMM. |
|
قیمت |
-
برای اعضای سایت : 100,000 Rial
-
برای دانشجویان عضو انجمن : 20,000 Rial
-
برای اعضای عادی انجمن : 40,000 Rial
|
خرید مقاله
|
|
.gif){kind=icon}