Speaker Name: | David Suendermann |
Speaker Info: | PhD student at Siemens Corporate Technology in Munich, Germany, and at the
Technical University in Barcelona, Spain (UPC) |
Date: | Thursday November 17 |
Time: | 11:30am-12:30pm |
Location: | CS Conference Room (Mudd) |
Abstract:
Residual prediction is a technique that aims at recovering the spectral details of speech that was encoded using parameterizations as linear predictive coefficients. Example applications of residual prediction are hidden Markov model-based speech synthesis or voice conversion. In this talk, I compare six residual prediction techniques and show that a novel method based on unit selection outperforms the others in terms of speech quality. |