2pNSc5. A computer-aided paired-comparison algorithm using a 5-point comparison scale and its application to performing subjective ratings.

Session: Tuesday Afternoon, June 17

Author: Hsiao-an Hsieh
Location: 15031 S. Commerce Dr., Rm. G308, Dearborn, MI 48120


The paired-comparison method (PCM) is a common tool used for sound quality studies in the automobile industry. In using PCM, evaluators are asked to compare two noise samples and indicate their preference for every possible sample--pair combination. A preference ranking of all samples is then generated. This paper describes the development of a PCM based on a five-point comparison scale. The primary motivations for this work were: (1) the need of predicting preference ranking for an individual or small group of evaluators, and doing so with reasonable repeatability and accuracy; and (2) predicting an individual's subjective acceptability ratings for all the samples without additional testing. Using a computer-optimization scheme on numerous simulated events, the author developed a weighting system for this comparison scale. A computer tool was subsequently written which incorporated this weighting system and measured the relative preference-distances among all noise samples based on an individual's PCM inputs. The preference ranking was readily available, and the predicted subjective ratings were obtained by mapping these measured distances on a linear scale of 1 to N (1=totally unacceptable, N=perfectly acceptable). Some successes of using this method on real-life noise samples are presented and discussed. [See NOISE-CON Proceedings for full paper.]

ASA 133rd meeting - Penn State, June 1997