METHOD FOR SPEAKER-INDEPENDENT PHONEME RECOGNITION IN A SPEECH SIGNAL Russian patent published in 2021 - IPC G10L15/187 

FIELD: computer technology.

SUBSTANCE: invention relates to the field of computer technology for processing audio data by information processing systems, and in particular to methods for constructing speech recognition systems. The effect is achieved by determining the frequencies of the formants in the sections of the speech signal and phonemic recognition of each section of the speech signal by comparing its phonetic features with the available data bank separately for each speech sound, where sequences of speech signals are formed from the speech signal, which are separated from the original signal for the period of the analyzed frequencies, the differences of the generated signals with the original signal are calculated, the levels of the difference signals in the analyzed area are calculated, the frequencies corresponding to the minima of the calculated levels are selected, the selected frequencies are grouped in pairs, and the pairs are grouped from the selected frequencies differing by at least 90 chalk, after which a pair is selected that has a minimum distance in the frequency plane F1, F2, where F1, F2 are the frequency axes corresponding to the studied range of the speech signal, from base pairs placed in the database, with the assignment of the analyzed area to the phoneme value of the base phonological pair.

EFFECT: increasing the reliability of speaker-independent speech recognition.

5 cl, 5 dwg