The goal of this project is to improve communication between different languages. We want to make non-native English speakers speak US English without collecting his/her English speaking data, and to make English speakers speak Chinese with his/her own voice. This project utilizes GMM voice conversion software.
The project will involve the collection of Chinese and English speech data in order to evaluate how well cross lingual voice conversion works. This will allow creation of a synthetic voice that speaks English built from the example of the speaker's Chinese data alone. The second part will involved taking existing English single speaker databases and making those spekaer speak Chinese but still sound like the original English speaker.
This work will utlize the CMU FestVox synthetic voice building toolkit.
Evaluation will be done by objective measures (simple speaker and language ID techniques) and through listening tests.