Subjective Evaluation of a High Quality Text-to-Speech System for Argentine Spanish

JORGE ALBERTO GURLEKIAN; CHRISTIAN COSSIO MERCADO; HUMBERTO TORRES; MARÍA VACCARI

Madrid

Conferencia; VII Jornadas en Tecnología del Habla and III Iberian SLTech Workshop, IberSPEECH 2012; 2012

ATVS Biometric Research Group of the Universidad Autónoma de Madrid, the Spanish Thematic Network on Speech Technology (RTTH) and the ISCA-Special Interest Group on Iberian Languages (SIG-IL).

This work summarizes the perceptual evaluation of our recently developed text-to-speech system (S1), based on unit concatenation. We compare it with two commercially available systems (S2 and S3) using three dierent evaluation methods. One is the P.85 recommendation by the International Telecommunication Union (ITU), the second method, called Syntactically Unexpected Sentences (SUS), is known to be the most strict for intelligibility evaluation, and the third is the Mean Opinion Score (MOS) scale. Results of ITU test showed better quality and intelligibility responses for system S2. General quality evaluated by MOS and intelligibility evaluated by SUS presented no appreciable dierences between S1 and S2. It is concluded that high quality performance is related to a complete intonation modeling of dierent type of phrase length and styles. S1 has high intelligibility and quality for general information sentences but presented lower scores for speci c tasks as proposed in ITU tests, where short phrases coverage should be introduced in the intonational modeling of our system.