Abstract:
It is well known that changes in vocal dynamic level do not result in a simple loudness increase. Rather, because the level of the higher harmonics tends to increase more rapidly than that of lower harmonics, distinct changes in timbre typically accompany the obvious changes in loudness. However, without reference to auditory transformations, it is difficult to determine which of the observed spectral changes are preserved in the ``auditory spectrum,'' or to estimate their perceptual effects. In order to address these issues, sung vowels were recorded from several classical singers. Data represent six voice classifications, three vowels, three dynamic levels, and three points within each singer's fundamental frequency range. Various unidimensional metrics based on models of auditory processes such as frequency selectivity, masking, loudness perception, and timbre perception [Glasberg and Moore, Hear. Res. 47, 103--138 (1990); Aures, Acustica 59, 130--141 (1985)] were obtained. Results to date indicate that the weighted first moments of the specific loudness and sharpness distributions serve as promising means of quantifying changes associated with changes in vocal dynamic level, and that these metrics are most effective when derived from excitation patterns based on auditory filter shapes rather than masking patterns. [Work supported by UGACOE Faculty Research Grant.]