Statistical identification of critical, dependent and redundant articulators

Jackson, PJB and Singampalli, VD (2008) Statistical identification of critical, dependent and redundant articulators

Full text not available from this repository.

Abstract

An automatic method for identifying critical, dependent and redundant roles in speech articulation is presented. Critical articulators are identified using the Kullback-Leibler divergence between phone-specific and model pdfs, which are initialised to the grand pdfs for each articulator. Model pdfs of critical and dependent articulators, those significantly correlated with the critical ones, are updated accordingly for both 1D and 2D cases, as long as the divergence exceeds the threshold. Those unaffected are termed redundant. Algorithm performance is evaluated on the MOCHA-TIMIT database by comparison with phonetic features. Results are also given for an exhaustive search, and principal component analysis of articulatory fleshpoints. Implications of being able to extract phonetic constraints automatically from articulatory recordings are discussed.

Item Type:	Conference or Workshop Item (UNSPECIFIED)
Divisions :	Surrey research (other units)
Authors :	Name Email ORCID Jackson, PJB p.jackson@surrey.ac.uk Singampalli, VD
Date :	July 2008
Depositing User :	Symplectic Elements
Date Deposited :	17 May 2017 12:48
Last Modified :	23 Jan 2020 18:06
URI:	http://epubs.surrey.ac.uk/id/eprint/836765

Actions (login required)

View Item

Downloads