Semantic Components of the Quality of Processed Speech

Publisher Website
Google Scholar
Abstract
Fifteen frequency distortions of one male voice were arranged on tape for paired-comparison presentation. Adult male and female listeners made quality-preference judgments and in addition rated each distortion on 30 semantic-differential scales. A paired-comparison scale of quality was determined and each of the SD scales was factored by means of an Eckart-Young type of procedure. This procedure proved sensitive to different points of view among the listeners, even when the differences concerned only one or two of the voice distortions. Results indicated that the semantic space for the distortions was two-dimensional, containing an Intelligibility Factor and a Naturalness Factor, and that the prediction of the quality scale required consideration of both of these factors.

This publication has 1 reference indexed in Scilit: