FIELD: technology for generating sounds similar to sounds of voice of healthy people using noise-like sounds of esophageal voice of people without a larynx.
SUBSTANCE: in accordance to the method, original signal is demodulated and resulting envelope curve is corrected and then multiplied by oscillation with transformed instant frequency, from input signal, signals of consonant sounds of speech are extracted as well as low frequency part of frequency range, which is demodulated, as well as spectral components adjacent to frequency of main voice tone, which are amplified, enhanced with signal, conjugated according to Gilbert, conjugated signals are multiplied, the signal, conjugated according to Gilbert and limited in amplitude is added to amplitude-limited signal resulting from multiplication, and multiplied with amplitude-limited signal which contains spectral components close to frequency of main tone of voice, signals of consonant sounds of speech and transformed signals of consonant sounds of speech are added and speech signal output is composed.
EFFECT: transformation of noise-like "vowel" sounds of esophageal voice to vowel sounds with discrete spectrum of harmonics, similar to the voice of a healthy person.
2 cl, 1 dwg
