Pathologic voice type and the acoustic prediction of severity

We hypothesized that acoustic measures would predict dysphonic severity with differential results for pathological voice types. An instructional program based upon synthesized voice signals was developed to facilitate an awareness of prototypical voice types. Eighty phonatory samples representing normal subjects as well as patients with unilateral vocal fold paralysis, vocal nodules, and functional dysphonia were analyzed acoustically on the basis of four measures: average fundamental frequency (F0), jitter, shimmer, and harmonic/noise ratio (H/N ratio). Following training, 29 listeners classified 62% of the phonatory samples on the basis of breathy, hoarse, rough, and normal. Dysphonic severity of rough voices was predicted more successfully by H/N ratio (r2 = .73) than by shimmer (r2 = .43). Dysphonic severity of breathy voices was predicted only by the combined features of less jitter, more shimmer, and lower H/N ratio (r2 = .74). No combination of acoustic variables was successful in the prediction of the hoarse voice type.