Extraction of phonemes from the merged speech and their identification

In this paper, we propose a method for determining the location of phonological formants that is not associated with the analysis of the amplitude spectrum and provides their isolation in the merged speech. The formants selected on this basis are grouped into phonemes, the correspondence of which to a particular sound is determined by the minimum distance from the base phonemes. A table of basic phonemes of vowels and several consonants is compiled. The regularity of the location of the basic phonemes and their binding to the coordinates of the location on the main membrane of the cortical organ is revealed. The results of the work are intended for creating autonomous automated speech recognition systems.