7th International Workshop on Models and Analysis of Vocal Emissions for Biomedical Applications (MAVEBA 2011)

Florence, Italy
August 25-27, 2011

Sentence Modality Recognition in Dysarthric Speech

D. Torres (1), T. Dekens (2), H. Martens (3), G. van Nuffelen (3), M. S. de Bodt (3), Werner Verhelst (2), C. A. Ferrer (1)

(1) Research Center on Electronics and Information Technologies (CEETI), Central University of Las Villas, Cuba
(2) Interdisciplinary Institute for Broadband Technology, Dept. of Electronics and Informatics (ETRO), Vrije Universiteit Brussel, Belgium
(3) Antwerp University Hospital, Rehabilitation Centre for Communication Disorders, Belgium

The ultimate goal of this research is to develop a tool for the automatic assessment and treatment of intonation and stress in dysarthric speech. In this paper, we deal with automatic sentence modality recognition in dysarthric speech. Two classes of sentence modalities were used: declarative statements and declarative questions. Statistics of prosodic features were used for the classification. Three well-known classification algorithms were tested with two different sets of features. The database used consisted of healthy and dysarthric speakers pronouncing three different sentences in both modalities. The healthy speakers were used as the training set and the dysathric speakers as the test set. A global classification accuracy of 84% has been achieved.

Index Terms. dysarthric speech, intonation, pitch, energy

Full Paper (reprinted with permission from Firenze University Press)

Bibliographic reference.  Torres, D. / Dekens, T. / Martens, H. / Nuffelen, G. van / Bodt, M. S. de / Verhelst, Werner / Ferrer, C. A. (2011): "Sentence modality recognition in dysarthric speech", In MAVEBA-2011, 71-74.