Multidimensional Perceptual Spaces of Strain and Roughness in Synthetic Female Singing Voices

Strain and roughness are psychoacoustic percepts evaluated during clinical voice assessments that commonly co-occur in dysphonic voices. The present study seeks to investigate if it is possible to disentangle inexperienced and experienced listeners' perception of strain and roughness through a paired-comparison task with multidimensional scaling analyses (MDS) and to identify acoustic correlates to emergent dimensions.

Synthetic female singing voices were generated at the pitches A3 and F5 on the vowel /ɑ/ with varying spectral source slopes to simulate varying levels of strain. A series of sine-wave amplitude modulations were superimposed on all synthetic voice stimuli to simulate varying levels of roughness. Stimuli were paired in all possible combinations within each pitch. Inexperienced and experienced listeners were asked to rate the dissimilarity of each stimulus pair using a visual-analog scale. MDS analyses were conducted to visualize the perceptual dimensions hypothesized to be associated with the perception of strain and roughness. Acoustic indices of first spectral moment, cepstral peak prominence, and local shimmer were calculated and entered into hierarchical linear regression analyses to identify predictors of each emergent MDS dimension.

Dimensions associated with the perception of strain and roughness emerged in all MDS solutions. Dimension 1 appeared to be associated with the perception of strain, and Dimension 2 appeared to be associated with the perception of roughness in all MDS solutions. Significant acoustic predictors emerged for each dimension.

Results from this study suggest that the perception of strain is separable from the perception of roughness. Additionally, acoustic predictors of the perception of roughness may vary depending on pitch. Results from this study validate the clinical use of perceptual voice assessments and have identified possible objective indices of strain and roughness.