Quantitative phonetic and frequency domain characterization of vocal blend for sung vowels

Abstract

Choral performance focuses on achieving the illusion of many voices singing as one smooth, cohesive tone, known as vocal blending. Choral pedagogy includes instructions on how to blend, but there is little information in the literature on how these qualitative instructions result in quantitative changes in the frequency domain. Voice samples of two altos with choral backgrounds were collected. They sung vowels [i], [u], [el, [o], and [[alpha]] together on a D4 (293 Hz), transitioning from unblended to blended tones, and data were analyzed using frequency domain techniques. For a musical pitch, the presence of higher harmonics amplifies dissonant intervals, which are undesirable in a blended choral sound. Choral singers are taught to ''darken" their tones (reduce the amplitude of higher harmonics) and modify their vowels by dropping their jaws and rounding their lips to achieve blend. The summed Fourier amplitudes for higher harmonics as a fraction of the sum of the fundamental frequency and second harmonic decreases by (82 ± 8)% for vowels [i] and [u], and (62 ± 12)% for vowels [e], [o], and [a] after transitioning to a blended sound. It can be concluded that significant overtone suppression occurs after blending. Vowel formants, the resonance frequencies of cavities in the vocal tract, amplify higher harmonics close in frequency to a given formant frequency. Across all vowels, the 2nd, 3rd, and 4th formant drop 1-3 harmonics between the unblended and blended time states, indicating that vowel modification reduces the intensity of dissonant intervals in higher harmonics by amplifying lower, more consonant harmonics.