Speech Synthesis by Using Single Phons as Speech Elements
W. Endres · The Journal of the Acoustical Society of America · 1974
As is known, continuous German speech can be synthesized using single phons, when connected by suitable transitional sounds. Recently we demonstrated that not only the /ə/, but also the /I/ and /ν/ ought to be used as transitional sounds. Their choice is dependent on the alleged phonetic environment. Additional investigations, using correlation methods in the time and frequency spheres, resulted essentially with the confirmation of this concept. The test material consisted of 196 German words spoken by one person. With a UNIVAC computer 1108, 16 characteristic normvowels of the German language were selected, and the degree of relationship between these normvowels-and then between the normvowels and the transitional sounds—was determined. As this synthesized speech should be used chiefly in automatic announcement and information services of the German Bundespost, it is important that the number of sound elements employed should be reduced. Simulated computer experiments, carried out in cooperation with the Heinrich-Hertz-Institut Berlin, showed that the character of certain vowels can be altered by shortening or lengthening their quasiperiodic parts. These results are demonstrated on tape.