Abstract
A computer model of a generator of complex periodic sounds simulating vowels has been developed. The system makes it possible to independently control each formant level and to immediately generate the signal. Spectrum envelope within the formant range is approximated by means of a trapezoid contour. A group of 6 listeners used the system to select synthesis parameters for sounds judged by each of them to be maximally similar to the Polish vowels. The sounds were subsequently subjected to evaluation by a more numerous panel of listeners, whereby sounds most suitable for prototypes of Polish vowels were selected. By controlled manipulation of the second and third formant levels, new sounds were obtained from the prototypes, which, after randomization, were presented for identification to seven listeners. The identification results obtained were presented in tabular form in three variants and characterized from the point of view of the needs of automatic recognition of vowels in continuous speech.References
[1] L.A. CHISTOVICH et al., Centers of gravity and spectral peaks as the determinants of vowel quality, Frontiers of Speech Communications Research (ed. B. Lindblom, S. Ohman), Academic Press, 145—157, 1978.
[2] G. FANT, Acoustic theory of speech production, Mouton, The Hague, 1960.
[3] G. FANT, et al., Two formant models, pitch and vowel perception, Auditory Analysis and Perception of Speech (ed. G. Fant, M.A.A. Tatham), Academic Press, 1975.
[4] T. HOUTGAST, Auditory analysis of vowel-like sounds, Acoustica, 4, 320—324 (1974).
[2] G. FANT, Acoustic theory of speech production, Mouton, The Hague, 1960.
[3] G. FANT, et al., Two formant models, pitch and vowel perception, Auditory Analysis and Perception of Speech (ed. G. Fant, M.A.A. Tatham), Academic Press, 1975.
[4] T. HOUTGAST, Auditory analysis of vowel-like sounds, Acoustica, 4, 320—324 (1974).
