Undergrad Research Project - Speech Synthesis Mimicking

Fall 2009

Ming Sun
Alan Black
Project description

Objective: Mimic system talks like the user.

Method: GMM voice conversion (mainly focus on pitch and duration)

Anticipated results: A machine shows user sentences to read after. It records the user's voice. Gradually, it mimics the speaker's speech style, speed and prosody to talk like them.

The Mimic System will be set up as an easy to use/demo application that will show off the current voice conversion technology.

The system will be evaluated with local users to find which aspect of voice conversion (F0/tune, duration and spectral) is perceived by the users as most relevant. This will be compared with objective measures in the conversion too.

Return to project list