Speech algorithms for Clinical and Educational Applications (SPACE): acoustical techniques for expressive specch synthesis.

Project Details

Description

Speech synthesis systems with good intelligibility and reasonable reliability are currently available for several languages. the best of these systems produce realisticread speech. However, for numerous classes of applications, such as language learning, security systems, humanoid robots, and computer aided learning, more expressive forms of speech are required. In an existing IWT-SBO project, we are developing a read speech synthesizer for Dutch that aims at extremely high quality for applications in computer assited training of dyslectic children. With this project, we wish to perform the required theoretical and experimental studies for providing this synthesizer with more expressive speech modi, such as encouraging, approving, disapproving, etc. We propose to perform experiments in which several speech parameters, such as pitch, loudness, timbre are transplanted from a database with expressive speech samples onto synthetic nonsense speech using a transplantation tool that was developed at our lab. Once we have experienced which speech parameters contribute to which expressive speech styles, machine learning techniques will be used to learn the dynamic behaviour of these speech features, such that the resulting rules can be incorporated in the synthesizer.
AcronymOZR1478
StatusFinished
Effective start/end date1/01/0731/12/08

Keywords

  • electricity

Flemish discipline codes

  • Other engineering and technology