Improving the Efficiency of a Dual Corpus Text To Speech Synthesis System Using a Prefetch Buffer

Abstract

 Text to speech synthesis systems use a corpus to convert any given arbitrary ext into the corresponding speech output. A corpus may contain any of the speech units such as phones, diphones, triphones, syllables etc.  Some synthesis systems use more than one corpus to improve the quality of the output speech, in which case, time taken to switch between both the corpus is more. This paper addresses this problem by efficient prefetching of speech units from both the corpus. The actual synthesis period is not disturbed.