A Syllable-Based Prominence Detection Model<br />Based on Discriminant Analysis and Context-Dependency

International audienceOn the basis of our previous work, we propose a syllablebased prominence detection model within the framework of exploratory data analysis and discriminant learning in the acoustic domain. This paper investigates two hypothesis on the acoustic data processing: a linear discriminant analysis in which the relative discriminant ability of single prosodic cues are combined into prosodic patterns and a context-dependant model that accounts for phonological dependencies (phonetic intrinsic properties and coarticulation effect). The proposed approach significantly outperforms a baseline method on a corpus of French read speech with a performance of 87.5% in f-measure for the prominent syllables (respectively 90.4% in global accuracy)