METHOD OF IDENTIFYING SPEAKER FROM ARBITRARY SPEECH PHONOGRAMS BASED ON FORMANT EQUALISATION Russian patent published in 2011 - IPC G10L15/00 

FIELD: information technology.

SUBSTANCE: identification of a speaker from arbitrary speech phonograms is carried out by evaluating similarity between a first phonogram of the speaker and a second reference phonogram. For the said evaluation, reference fragments of speech signals are selected on the first and second phonograms, on which there are formant tracks. At least three formants compare reference fragments in which values of at least two formant frequencies coincide. The similarity of the compared reference fragments is evaluated on the coincidence of values of the rest of the formant frequencies, and similarity of the phonograms on the whole is determined from the overall evaluation of similarity of all compared reference fragments.

EFFECT: reliable identification of a speaker for long and short phonograms, phonograms recorded in different channels with high level of interference and distortions, as well as phonograms with arbitrary speech of speakers in different psychophysiological states, speaking in different languages.

6 cl, 8 dwg