Computer Simulation of a Formant-Vocoder Synthesizer
James L. Flanagan, Cecil H. Coker, Carol M. Bird · The Journal of the Acoustical Society of America · 1962
A speech synthesizer of the formant-vocoder type is described in which vocal antiresonances (zeros) as well as resonances (poles) are taken into account. The synthesizer is simulated by sampled-data operations on a large-scale digital computer. Digital-to-analog conversion of the output permits listening to the synthetic sounds. Voiced sounds are generated by periodic (glottal) excitation of one branch of the circuit. Transfer functions for vowel and vowel-like sounds are described by 4 poles. Nasal consonants are represented by 5 poles and 1 zero. Voiceless sounds are generated by noise excitation of a separate branch. Fricative consonants are described by 1 zero and either 1 or 2 poles. Frequencies and bandwidths of the poles and zeros are independently controllable, although certain vocal constraints are capitalized upon to relate their values. Amplitude of voiced and voiceless excitation and pitch are also controlled. Performance of the synthesizer with band-analyzed control data is demonstrated. Special computer input equipment for facilitating automatic extraction of the control data from real speech is also described.