Spanish Emotional Speech Synthesis

Abstract

In this project a text-to-speech (TTS) HMM-based speech system (HTS) has been used to create emotional synthetic speech in Spanish. Nowadays the synthetic voices have high quality, but this is not enough, they must be able to capture the natural expressiveness of the human speech. Giving this expressiveness to the synthetic voices will lead to a much more natural voice, that is the goal of these systems.To achieve this, both male and female voices will be used and two different techniques will be applied: dependent models and average voice models with adaptation.In this TTS system diffeerent vocoders can be used. For this project GlottHMM has been used and then three perceptual test have been carried out to compare it with STRAIGHT vocoder.The results of the perceptual tests shows that STRAIGHT is very robust and that GlottHMM is not yet at its level regarding the emotional speech synthesis.