AN AUDITORY-BASED DISTORTION MEASURE WITH APPLICATION TO CONCATENATIVE SPEECH SYNTHESIS

This study presents a new auditory-based distance measure with applica tion to concatenative speech synthesis. This measure employs the Carne y auditory model to produce a feature vector related to auditory perce ption. For concatenative synthesis, the new measure is employed to ass ess perceived discontinuities at segment transitions. Evaluations usin g a restricted data base environment show that the new measure can be effective in improving speech synthesis performance.