Second ESCA/IEEE Workshop on Speech Synthesis
September 12-15, 1994
Concatenating speech synthesis units give rise to temporal discontinuities in the short time spectral envelope of a synthetic speech signal. Among the various methods for increasing the speech fluency, three methods for unit segmentation and two methods for interpolation were applied to our demisyllable-based TD-PSOLA synthesizer. By means of a pair-comparison test it was found that no significant quality increase could be achieved by any of the evaluated concatenation methods.
Bibliographic reference. Kraft, Volker (1994): "Does the resulting speech quality improvement make a sophisticated concatenation of time-domain synthesis units worthwhile?", In SSW2-1994, 65-68.