Acoustic identification of Japanese vowels and its application to pathological articulation

Abstract

This study acoustically investigated Japanese vowels, and three acoustic cues were found. The first acoustic cue is the ratio of the second formant frequency to the first formant frequency. The second cue is the ratio of the third formant frequency to the second formant frequency. Both of these cues can be explained theoretically using perturbation theory for a singular acoustic tube model. The third acoustic cue is determined by the relative strength in the intermediate region between the second and third formant frequencies and the threshold value. This cue is related to the auditory masking phenomenon. We determined three criteria for these three acoustic cues based on the voices of 84 subjects. We found that the problem of speaker normalization of vowels in speech perception was solved by taking the third acoustic cue into account. These findings confirmed that this method is appropriate for evaluating pathological articulation and can be applied to acoustic analysis of pathological voices.