Evaluating models of vowel perception

Abstract

There is a long-standing debate concerning the efficacy of formant-based versus whole spectrum models of vowel perception. Categorization data for a set of synthetic steady-state vowels were used to evaluate both types of models. The models tested included various combinations of formant frequencies and amplitudes, principal components derived from excitation patterns, and perceptually scaled LPC cepstral coefficients. The stimuli were 54 five-formant synthesized vowels that had a common F1 frequency and varied orthogonally in F2 and F3 frequency. Twelve speakers of American English categorized the stimuli as the vowels /I/, /[symbol: see text]/, or /[symbol: see text]/. Results indicate that formant frequencies provided the best account of the data only if nonlinear terms, in the form of squares and cross products of the formant values, were also included in the analysis. The excitation pattern principal components also produced reasonably accurate fits to the data. Although a wish to use the lowest-dimensional representation would dictate that formant frequencies are the most appropriate vowel description, the relative success of richer, more flexible, and more neurophysiologically plausible whole spectrum representations suggests that they may be preferred for understanding human vowel perception.

Identification of high-pass filtered male, female, and child vowels: The use of high-frequency cues

“…These results would not be predicted based on current formant models of vowel perception, which focus primarily on information in the frequency region that was removed from the vowel signals used in the current study. As previously noted, research to date has focused on lower frequency spectral resonances when describing vowels and examining vowel identification performance (Peterson and Barney, 1952;Lehiste and Peterson, 1959;Miller, 1989;Hillenbrand et al, 1995;Molis, 2005). In contrast, the present results provide support for the study of perceptual information contained in the high-frequency region of vowels.…”

“…Vowels are characteristically described according to low-frequency resonance characteristics (e.g., Peterson and Barney, 1952;Hillenbrand et al, 1995;Molis, 2005;Swanepoel et al, 2012). Classically, the study of vowel perception has focused on the lowest formant frequencies, typically F1, F2, and F3.…”

The influence of noise on vowel and consonant cues

“…Since the classic study by Peterson and Barney ͑1952͒ on the distribution of the vowel formant frequencies on the F1-F2 plane, many studies were conducted in quiet on vowel perception ͑see reviews by Strange, 1989 of difference limens for formant discrimination ͑Flanagan, 1955; Hawks, 1994;Liu and Kewley-Port, 2001͒, vowel modeling ͑e.g., Syrdal and Gopal, 1986;Molis, 2005͒, and other aspects of vowel discrimination. Many factors were found to be important for vowel identification including formant frequencies ͑Peterson and Barney, 1952͒, vowel duration ͑e.g., Ainsworth, 1972͒, F0 ͑e.g., Lehiste andMeltzer, 1973͒, spectral contrast ͑Leek et al, 1987;Poroy, 2001͒, formant contour ͑Hillenbrand andGayvert, 1993;Hillenbrand and Nearey, 1999͒, spectral shape ͑e.g., Zahorian and Jaghargi, 1986;Ito et al, 2001͒, and spectral change ͑e.g., Strange et al, 1983͒. Of these factors, the formant frequencies and spectral shape, in particular, have been found to be major cues to vowel perception.…”

The relative importance of spectral cues for vowel recognition in severe noise

“…At the most severe noise level, À10 dB SNR, vowels containing detailed information on the spectral shape were recognized significantly better than vowels consisting only of formant information. Despite the fact that formant cues alone have been found to be adequate for vowel perception in quiet (Delattre et al, 1952;Klatt, 1982;Molis, 2005), a more complete description of the vowel spectrum in severe noise conditions resulted in better recognition. Zahorian and Jagharghi (1993) proposed that the whole-spectrum representation of a vowel provides some relevant spectral information that is not provided by formants.…”