Smooth ergodic hidden Markov model and its applications in text to speech systems

Abstract

In text-to-speech "ystems. the (/CCI/IYICY of information extraction Fom text is crucial iii producing high quality synthesized speech. In this paper, a nell' scheme for conl'erling lext into ils eqllil'alent phonetic spelling is proposed {lnd del 'eloped_ This method has many advantages over its predecessors and it can complement muny other text to speech converting systems ill order to get improved pe/torn/ance.