Does vowel recognition relate to pitch?
Acoustics Virtually Everywhere (AVE), Maurer et al., 2020


> Close

Background


The relation between recognised vowel quality and the vowel spectrum is nonuniform.


Sound series for illustration
 

The relation between fo, spectral peaks and spectral envelopes is nonuniform in terms of the occurrence of:
    sounds with different numbers of vowel-related spectral peaks
    sounds with “flat” or sloping vowel-related spectral portions
    absence of fo-dependence of vowel spectra and formant patterns and, consequently,
    absence of formant pattern and spectral shape ambiguity for fo < c. 200–250 Hz for
    all vowels   
    cases of absence of formant pattern and spectral shape ambiguity for a substantial part
    of sounds of the vowels /ɛ–a/

These aspects were previously documented (see below, References). Here, as examples, two aspects are illustrated (excerpt of Maurer et al, 2019):
    absence of fo-dependence of vowel spectra for fo < c. 200–250 Hz
    cases of absence of formant pattern and spectral shape ambiguity for sounds of the
    vowel /a/
   

Role of frequency levels and ranges of fo – absence of fo-dependence of vowel spectra for fo < c. 200–250 Hz

As mentioned and illustrated in the previous paragraphs, vocalises (vowel sounds produced with stepwise rising fo) above all of closed and mid-closed vowels indicate a substantial relation between fo and the lower spectrum for fo variation above 200–250 Hz, effecting formant pattern and spectral shape ambiguity of vowel sounds. However, this is not the case for fo variation below 200–250 Hz.

In these terms, spectral representation of vowel quality is nonuniform concerning fo variation. This is illustrated in the following series.


Series 1. Two sounds of /i/ produced by a man, with fo variation of 107–240 Hz (approximately one octave). No pronounced change in the lower spectrum with increasing fo is manifest.

> Sounds

Series 2. Two sounds of /i/ produced by a woman, with fo variation of 266–524 Hz (approximately one octave). A pronounced change in the lower spectrum with increasing fo is manifest.

> Sounds


Series 3. Two sounds of /ø/ produced by a man, with fo variation of 100–198 Hz (approximately one octave). No pronounced change in the lower spectrum with increasing fo is manifest.

> Sounds

Series 4. Two sounds of /ø/ produced by a woman, with fo variation of 174–345 Hz (approximately one octave). A pronounced change in the lower spectrum with increasing fo is manifest.

> Sounds


Role of vowel qualities

As also mentioned, the sounds for the closed and mid-closed vowels illustrate a pronounced relation of the lower vowel spectrum and fo, but the sounds for the mid-open and open vowels illustrate inconsistent or weak or absent relation of the lower vowel spectrum and fo and point at the nonuniform character of the relation.

In these terms, spectral representation of vowel quality is nonuniform among different qualities.  This is illustrated in the following series.


Series 5. Four sounds of the closed vowel /i/ produced by women in the fo range of 150–635 Hz, with pronounced change in the lower spectrum.

> Sounds

Series 6. Four sounds of the mid-closed vowel /e/ produced by women in the fo range of 195–573 Hz, with pronounced change in the lower spectrum.  

> Sounds

Series 7. Four sounds of the open vowel /a/ produced by women in the fo range of 169–664 Hz, with no indication of a change in the lower spectrum.  

> Sounds



The sounds were previously shown in Maurer et al (2019). For additional illustration, see that publication:
 
Maurer, D., Suter, H., d'Hereuse, C., & Dellwo, V. (2019). Formant Pattern and Spectral Shape Ambiguity of Vowel Sounds, and Related Phenomena of Vowel Acoustics-Exemplary Evidence. In INTERSPEECH (pp. 2368-2369).

> Paper
> Online presentation