The phonetic database

Presentation of speech database elements specified with phonetic data

Hungarovox, and Multivox Hungarian speech synthesizers used such data, 1982-1990.

The phonetic database contains a set of short “audio” slices characterised by phonetic data.

The phonetic data per slice in the database are as follows:

  • F1, F2, F3, F4 formant frequency value (Hz)
  • Value of formant bandwidth B1, B2, B3, B4 (Hz)
  • Pitch - PI parameter and Hertz
  • Volume (dB) - AM
  • Excitation (voiced = V; noisy = c)
  • Duration = 12ms / slice - FD

An example of concatenating a series of paramter slices used for the synthesis of the first syllable of the Hungarian expression “Good morning” (Jó napot). (1-4 slices = j); (5-9 = j-o sound transition); (9-17 = o )

alter-text

The entire database contained 370 types of parameter slices, with which all the speech sound connections of Hungarian speech could be realized. Memory requirement was 1 Kbyte. The sound of the synthesized speech was robotic, but well understood.