Sixth European Conference on Speech Communication and Technology

Budapest, Hungary
September 5-9, 1999

Acoustical Features as Predictors for Prominence in Read Aloud Dutch Sentences Used in ANN's

Barbertje M. Streefkerk (1), Louis C. W. Pols (1), Louis F. M. ten Bosch (2)

(1) Institute of Phonetic Sciences Amsterdam (IFA)/IFOTT, The Netherlands
(2) Lernout & Hauspie Speech Products N.V., Belgium

In this paper we present several acoustical features, which are used as predictors for prominence. A set of 1244 sentences from 273 different speakers is selected from the Dutch Polyphone Corpus. Via listening experiments the subjective prominence markers are obtained. Several acoustical features concerning F 0 , energy and duration are derived and used as predictors for prominence. The sentences are divided in a test and a training set, to test and train neural networks with different topologies and different input features. The first results show that a classification of prominent and non-prominent words is possible with 82.1% correct for an independent test set.

Full Paper (PDF)

Bibliographic reference.  Streefkerk, Barbertje M. / Pols, Louis C. W. / Bosch, Louis F. M. ten (1999): "Acoustical features as predictors for prominence in read aloud dutch sentences used in ANN's", In EUROSPEECH'99, 551-554.