Abstract
180 isolate voicings of (near) Cardinal Vowels - 10 of each: [i e ε a y ...] were described using four formant frequencies, which were measured from FFT spectra. In a 4-D space, the tokens could be correctly assigned 95% of the time using Bayes estimators of discriminant scores. The error increases with the reduction of the number of variables to three or two. The mean vectors characterizing the 18 vowel classes were, for the present speaker, somewhat different from those recorded and measured about 20 years ago, which may be due to aging.References
[1] W.A. AINSWORTH, Intrinsic and extrinsic factors in vowel judgement in: Auditory analysis and perception of speech, [Eds.] G. Fant, M.A.A. Tatham, Academic Press, London 1975, 103-116.
[2] L. ARMSTRONG, D. JONES, The phonetics of French, G. Bell and Sons Ltd., London 1951.
[3] A. BUTCHER, Cardinal Vowels and other problems, in: Linguistic controversies, [Ed.] D. Crystal, E. Arnold, London 1982, 50-71.
[2] L. ARMSTRONG, D. JONES, The phonetics of French, G. Bell and Sons Ltd., London 1951.
[3] A. BUTCHER, Cardinal Vowels and other problems, in: Linguistic controversies, [Ed.] D. Crystal, E. Arnold, London 1982, 50-71.
