Abstract: This paper describes the GTHCSTR entries developed for the Speech Synthesis Albayzin 2010 Evaluation. We have developed three different HMM-based systems to build synthetic voices in Spanish using two hours of speech of a male speaker. We have improved our baseline system (GTHCSTR-2008) by using morphosyntactic features, iterative segmentation algorithms, enhanced feature analysis and speaker adaptation techniques.
Index Terms: text to speech synthesis, statistical parametric speech synthesis, morphosyntactic features, speaker adaptation.