Speech Synthesis with Prerecorded Syllables and Words

Abstract

Lists of Psycho-Acoustic Laboratory sentences were used for synthesis of speech with prerecorded words and syllables as the building blocks. The sentences were phonemically transcribed with syllable boundaries indicated. The syllables were spoken in a random order with pauses between each syllable. The words were spoken in the same way without being phonemically transcribed. Two rates, slow and fast, were used for each type of building block, making four basic conditions. A different group of eight listeners heard each condition of synthesized speech. The mean scores, percent correct of the five key words in each sentence, for each condition were: fast rate syllables, 66.6%; fast rate words, 60%, slow rate syllables, 90%; slow rate words, 95.6%.