Experimental Setup for the Synthesis of Speech Segments

摘要

The report describes an experimental setup for synthesizing speech for the purpose of accumulating data on the structural make-up of the discrete units of language and of determining the technological demands made on such devices. The stream of speech at the synthesizing end of the system is regenerated under the immediate control of the portion of information abstracted from the speech of the announcer. As applied to reading and translating machines the problem of synthetic speech appears in a somewhat different form: sufficiently intelligible artificial speech must be created anew from the semantic information obtained in the reading or translation of some text. A block-diagram is given whose principal components are a sound power source, unit for forming a statistical formant spectrum, units for formant variation, automatic starter for voicing the synthesized speech segment, delay block, electronic commutator for pauses between sounds, electronic channel commutators, device for shaping the synthesized speech segment, and final amplifier. The described device synthesized separate phonemes, words, and syllables of Georgian. Videograms of synthesized speech segments and brief engineering data on the individual components of the setup are given.