Predicting Unseen Articulations from Multi-speaker Articulatory Models

• In order to study inter-speaker variability, this work aims to assessthe generalization capabilities of data-based multi-speakerarticulatory models. We use various three-mode factor analysistechniques to model the variations of midsagittal vocal tractcontours obtained from MRI images for three French speakersarticulating 73 vowels and consonants. Articulations of agiven speaker for phonemes not present in the training set arethen predicted by inversion of the models from measurementsof these phonemes articulated by the other subjects. On the average,the prediction RMSE was 5.25 mm for tongue contours,and 3.3 mm for 2D midsagittal vocal tract distances. Besides,this study has established a methodology to determine the optimalnumber of factors for such models.

Ämnesord

NATURVETENSKAP  -- Data- och informationsvetenskap -- Datavetenskap (hsv//swe)

NATURAL SCIENCES  -- Computer and Information Sciences -- Computer Sciences (hsv//eng)

NATURVETENSKAP  -- Data- och informationsvetenskap -- Språkteknologi (hsv//swe)

NATURAL SCIENCES  -- Computer and Information Sciences -- Language Technology (hsv//eng)

Nyckelord

Factor analysis

Multi-speaker articulatory model

Publikations- och innehållstyp

ref (ämneskategori)

kon (ämneskategori)