FIELD: physics.
SUBSTANCE: value of the period of the fundamental tone is determined in each frame of the speech signal. The vector of the values of the fundamental tone is stored in form of signal vectors which are then used to determine the error functions of time intervals. The average spectrum of time interval errors is then calculated by averaging the amplitude of spectral components of the spectra of error functions of the time intervals of neighbouring signal vectors. An array of frequencies FP is created, which corresponds to frequencies of peak amplitude values in the average spectrum of time interval errors. A modified spectrum of periodic jitter is calculated on frequencies FP with amplitude values of the average spectrum of time interval errors and phase values of the spectrum of time interval errors. The time sequence of the periodic jitter is calculated via inverse Fourier transformation from the modified spectrum of the periodic jitter. The modified spectrum of random jitter is calculated by substitution on frequencies FP of the amplitude, equal to the square root of the difference between the average spectral density of power and the square of the modulus of the average spectrum of time interval errors, and incidentally, uniformly distributed phase.
EFFECT: method of dividing jitter of fundamental tone of speech signal into random and periodic components using a spectral technique.
3 dwg
Authors
Dates
2011-05-20—Published
2009-12-01—Filed