Identification of speakers by use of nasal coarticulation

Abstract

A new approach, based on the statistical properties of the nasal spectra, has been used to quantitatively study the coarticulation of nasal consonants, with the vowels following them in isolated /hə′CVd/ utterances. The spectral differences between the mean spectra of nasals followed by front vowels and those of nasals followed by back vowels were used as the acoustic measure of the coarticulation of [m] and [n] with the following vowel [V]. The coarticulation between [n] and [V] was found to be only about one-third of that between [m] and [V]. The coarticulated nasal spectrum, particularly between [m] and [V], was found to have strongly idiosyncratic characteristics, which are not likely to be modified in natural speech. A method was therefore developed by which the coarticulation between [m] and [V] was taken as the acoustic clue and the speaker was identified by use of a correlation decision criterion. Coarticulation was found to give more reliable clues than the nasal spectrum alone, which had earlier been found to be one of the best acoustic clues for identifying speakers.