An improved correction formula for the estimation of voice source harmonic magnitudes

Abstract

Information about voice and talker characteristics, or voice quality, can be gained from glottal flow estimates. Many voice quality parameters, such as the open quotient (OQ), depend on an accurate estimate of the glottal flow (voice source) spectrum. It is known that OQ, for example, is correlated with the magnitude difference of the first two harmonics (H1–H2) of the voice source spectrum [Holmberg et al., ‘‘Comparisons among aerodynamic, electroglottographic, and acoustic spectral measures of female voice,’’ J. Speech Hear. Res. 38, 1212−1223 (1995)]. To obtain an accurate estimate of the voice source spectral harmonics, the influence of vocal−tract resonances needs to be removed. In Hanson, Ph.D. dissertation, Harvard University (1995), a correction formula which removes the effect of F1 on H1 and H2 was presented. In this work, an improved correction formula for the estimation of source harmonics′ magnitudes is presented. The new correction formula accounts for the bandwidths of vocal−tract resonances, and most importantly, is not limited to the analysis of nonhigh vowels. H1−H2 estimates, using the proposed technique with synthesized vowels /a/, /i/, and /u/ generated with the LF and the KLGLOTT88 models, are very accurate. [Work supported in part by the NSF.]