Spectral features related to the auditory perception of twang-like voices

To the best of our knowledge, studies on the relationship between spectral energy distribution and the degree of perceived twang-like voices are still sparse. Through an auditory-perceptual test we aimed to explore the spectral features that may relate with the auditory-perception of twang-like voices.

Ten judges who were blind to the test's tasks and stimuli rated the amount of twang perceived on seventy-six audio samples. The stimuli consisted of twenty voices recorded from eight CCM singers who sustained the vowel [a:] in different pitches, with and without a twang-like voice. Also, forty filtered and sixteen synthesized-manipulated stimuli were included.

Based on the intra-rater reliability scores, four judges were identified as suitable to be included in the analyses. Results showed that the frequency of F₁ and F₂ correlated strongly with the auditory-perception of twang-like voices (0.90 and 0.74, respectively), whereas F₃ showed a moderate negative correlation (-0.52). The frequency difference between F₁ and F₃ showed a strong negative correlation (-0.82). The mean energy between 1-2 kHz and 2-3 kHz correlated moderately (0.51 and 0.42, respectively). The frequency of F₄ and F₅, and the energy above 3 kHz showed weak correlations. Since the spectral changes under 2 kHz have been associated with the jaw, lips, and tongue adjustments (i.e. vowel articulation) and a higher vertical laryngeal position might affect the frequency of all formants (including F₁ and F₂), our results suggest that vowel articulation and the laryngeal height may be relevant when performing twang-like voices.