Isopreference method for speech evaluation.

A study of speech‐quality measurement procedures requires a working definition of the over‐all concept of speech quality. It is assumed that speech quality can be described in terms of intelligibility, preference, loudness, and speaker recognizability. A utilitarian approach to the measurement of preference, the isopreference method, is presented and discussed. In this method, preference is evaluated in a forced pair‐comparison test and is described in terms of a continuously variable reference signal. The results of extensive testing and test repetitions indicate that the accuracy of preference evaluations is not limited by the proposed method, but by variety and variability of listeners and by training and accommodation effects for given sets of test signals.