[Perceptual assessment of dysphonia: a training protocol with natural speech]

In this study, we proposed and tested the effectiveness of a training procedure on perceptual evaluation of dysphonia. Participants were naive listeners who learned to judge the severity of vocal disorders.

The corpus consisted of 142 female voices ranging from normal to severely dysphonic voices. A total of 38 naive listeners were tested, and the experiment was divided in 3 phases: (1) A pretest to assess the level of our listeners in the categorization of dysphonia, (2) the training phase and (3) a post-test to assess the improvement associated with the training. To examine the persistence of the training, the post-test was administrated twice: Immediately after the training session and one week later.

Improvement in the performance between the pretest and the post-test was found for moderate and severe dysphonia. No improvement was observed in the categorization of slightly dysphonic voices, which seem the more resistant to learning. Normal voices also showed no improvement, which is likely due to the high performance on these voices during the pre-test. The improvement observed in the categorization of moderate and severe dysphonia was still present even one week after the training.

The protocol tested in this study appears to be effective and could constitute an element of training courses for speech and voice therapists. The perception of the intermediate levels of dysphonia, however, remains difficult to quantify and needs to be improved.