Inexactness and robustness in cepstral-to-formant transformation of spoken and sung vowels

Abstract

Accurate measurement of formant frequencies is important in many studies of speech production and perception. The mapping of formant frequencies from cepstra coefficients by Broad and Clermont (1989) is an important evolution in formant estimation techniques from previous studies. This method also holds significant potential in estimating formant frequencies of sung vowels due to its improved robustness over traditional formant tracking techniques. The employment of this method in the estimation of formant frequencies from a dataset of spoken and sung phonation has led to some findings, not only on the inexactness of the model, but also on the resonant nature of the sung and spoken vowels themselves.