Method of and system for determining the pitch in human speech

In a system of speech analysis used, for example, in a vocoder in which the pitch value of human speech is determined and subsequently transmitted together with other speech parameters to the receiving section of another vocoder wherein the speech is synthesized and reproduced, a method for determining the pitch of a speech signal comprising the steps:

• analyzing the amplitude of said speech signal by regularly selecting time segments of the speech signal, determining from each time segment a sequence of spectrum components which constitutes the discrete Fourier transform of samples of speech signal, and deriving in each time segment the position of the significant peaks in the spectrum from the sequence of spectrum components;
• selecting an initial value for the pitch and determining a sequence of consecutive integral multiples of this pitch value, harmonic numbers corresponding to the multiplication factor in said multiples being associated respectively with said multiples of said pitch value;
• establishing intervals around the significant peak positions, said intervals defining a mask having apertures in situ of said peak positions;
• determining the number of multiples of said pitch value which coincide with said mask apertures;
• computing a quality figure in accordance with a criterion indicating the degree to which said multiples of said pitch value and said mask apertures match;
• repeating said immediately preceding determining and computing steps using consecutively higher values of pitch and multiples thereof, as determined in said selecting step, up to a predetermined highest value, resulting in the computing of separate quality figures associated with each of said pitch values;
• selecting the value of pitch having the highest associated quality figure and designating this pitch value as a reference pitch value;
• associating the harmonic numbers of the multiples of the reference pitch value with the significant peak positions located in the same aperture, thereby defining the location of said significant peak positions in a sequence of harmonics of the same fundamental tone;
• determining a probable value for the pitch of the speech signal, wherein the deviations between the significant peak positions and the corresponding multiples of the probable value having the same harmonic numbers are as small as possible;
• and
• combining said determined pitch value with other speech parameters for subsequent transmission or storage thereof in, for example, a read-only-memory.