अमूर्त

Computational Study of Fundamental Frequency of Standard Yor�b� monosyllabic Utterances

Michael Adébísí Fáyẹ̀míwò and Ọdẹ́túnjí Àjàdí Ọdẹ́jọbí

This paper presents the development and evaluation of a computational model for the fundamental frequency (F0) of Standard Yorùbá (SY) monosyllabic utterances. This was done with a view to approximating the F0 curves on SY syllables in the context of speech applications and other speech technologies. A list of 39 SY syllables which are the most frequently occurring syllables in Yorùbá newspapers and textbooks was compiled and the speech sound corresponding to the selected SY syllables were recorded for 5 adult native male speakers of SY. The F0 of the speech data were extracted using Praat speech processing package. Thereafter, Least Square Method (using polynomial degree of 1 to 7) was used to design the computational model for the F0 patterns extracted. The computational model was evaluated using the quantitative and the qualitative techniques. The results showed that the 7th degree polynomial had the lowest RMSE value for monosyllabic speech data. The results of the model showed that polynomials with degree 3 gave a good approximation for monosyllabic. The study thereafter established the computational and perceptual correlations between the F0 curves and the three Yorùbá tones (High, Mid and Low) for SY monosyllabic utterances. The modeling of F0 contour for Yorùbá tones for bisyllabic utterances and continuous speech are the areas of further research works, in which the principle of this work could be extended.