OPPapers.com Essay Index >> Technology >> High-Level Speaker Verification Via Articulatory-Feature Based Sequence Kernels And Svm
We have many free term papers and essays on High-Level Speaker Verification Via Articulatory-Feature Based Sequence Kernels And Svm. We also have a wide variety of research papers and book reports available to you for free. You can browse our collection of term papers or use our search engine.
High-Level Speaker Verification via Articulatory-Feature based Sequence Kernels and SVM begin{abstract}vspace{-0.06cm} Articulatory-feature based pronunciation models
Submitted by talentboy on April 29, 2008
Category: Technology
Words: 4145 | Pages: 17
Views: 75
Popularity Rank: 112,908
Average Member Grade: N/A (Add a Comment / Grade this Paper)
\begin{abstract}\vspace{-0.06cm}
Articulatory-feature based pronunciation models (AFCPMs) are capable of
capturing the pronunciation variations among different speakers and are good
for high-level speaker recognition. However, the likelihood-ratio scoring
method of AFPCMs is based on a decision boundary created by training the target
speaker model and universal background model (UBM) separately. Therefore, the
method does not fully utilize the discriminative information available in the
training data. To fully harness the discriminative information, this paper
proposes training a support vector machine (SVM) for computing the verification
scores. More precisely, the models of target speakers, individual background
speakers, and claimants are converted to AF-supervectors, which form the inputs
to an AF-based kernel of the SVM for computing verification scores. Results
show that the proposed AF-kernel scoring is complementary to likelihood-ratio
scoring, leading to better performance when the two scoring methods are
combined. Further performance enhancement was also observed when the AF scores
were combined with acoustic scores derived from a GMM-UBM system.
%However, to represent the impostor population, the likelihood-ratio scoring
%method of AFPCMs only uses a single universal background model (UBM) that is
%trained without considering the target speakers; therefore this scoring method
%does not fully utilize the discriminative information available in the training
%data.
\end{abstract}
%\noindent{\bf Index Terms}: Speaker verification, kernels, articulatory
%features, pronunciation models, SVM \vspace{-0.1cm}
\section{Introduction}\label{sec:intro}%\vspace{-0.1cm}
Studies have shown that combining low-level acoustic information with
high-level...
You must Login to view the entire paper.
If you are not a member yet, Sign Up for free!