Isopreference Method for Evaluating Speech-Transmission Circuits

Publisher Website
Google Scholar
Abstract
This exploratory paper describes a modification of the paired comparison technique for deriving a one-dimensional scale for rating speech transmission systems on the basis of listener preferences. The numbers on the scale, which run from 0 to 100, are called “Transmission Preference Units” (TPU), and are intended to be used to evaluate any speech transmission system, regardless of the noise or distortions encountered, provided the system is less preferred than the reference condition, namely, real speech at 1 m. If the TPU ratings for two transmission systems are known, it is believed that the difference can be used to predict the percentage of users who would prefer the system with the higher rating. Subjective listening experiments have been conducted with a small group averaging seven observers for rating a large number of speech transmission conditions on the TPU scale. Equal preference contours were determined on speech level-noise level planes for several different frequency bands. The listening tests require only simple A-B preference comparisons of two brief samples of speech heard consecutively. The data suggest that the basic requirement of transitivity, essential for a simple useful one-dimensional rating scale, may be satisfied for many types of transmission conditions. Thus if transmission condition A is judged to equal condition B in preference, and condition B equals C, then A and C were also found to be judged equal. This apparently holds even though A, B, and C are circuits differing widely in their physical parameters. This study is reported at this stage in the hope that other laboratories may become interested in checking and extending the limited scope of the work reported here. Such confirmation is necessary before a simple one-dimensional TPU scale can be postulated and used with confidence for precise evaluations. For the purposes of this work, two circuits are considered to be equal in preference when half of the judgments of the group favor one and half favor the other circuit; it does not follow that individual observers are indifferent in their preferences for the two circuits.

This publication has 0 references indexed in Scilit: