Short description

Electro-Larynx (EL) speech sounds very robotic. This is, because – among other reasons – the voice source is designed only to maximize intelligibility and not to meet any naturalness criteria.

The aim is to design a sound source synthesizer, that runs in real-time and parameters can be modified online. This should be done in order to be able to introduce variation in the sound, such a pitch contour, timbre, etc.

The student can build on a previous diploma thesis which looked at several voice source generation algorithms

Your Tasks

• Review of literature
• Implementation using Matlab
• Intelligibility and speech quality evaluation of produced speech

Your Profile/Prerequisites

• Speech Communication 1
• Speech Communication Laboratory is recommended
• Matlab

Contact:

Martin Hagmüller (hagmueller@tugraz.at or 0316/873 4377)