A Comparison of Preference Measurement Methods

Abstract

The present paper compares four methods for measuring the speech-quality parameter preference. The scales of the isopreference method, the category judgment method, the relative preference method, and the absolute preference judgment method (a proposed modification of the category judgment method) are related to a “preference unit” scale. Capabilities and performance of the methods are discussed and illustrated by evaluating a set of test signals with different types of degradation containing a subset of clearly discriminable signals. The paper tries to show how far preference results gained with one method allow prediction of corresponding results in terms of another method. Because of the quantitative limitations of subjective tests and especially since the four methods use different approaches and have different application ranges, it is not unexpected that the preference unit scale has been found to be more of scientific rather than of practical engineering interest. Listeners have been found capable of discriminating reliably, even without reference or anchor signals, more than the five quality steps provided by the category judgment method, and somehow also by the relative preference method. The proposed absolute preference judgment method recognizes this finding and, therefore, yields much better agreement with the isopreference method than the two other methods, which are apparently too strongly quantized. © 1971, Acoustical Society of America. All rights reserved.