Assessments of vehicle interior noises—Considerations about the quality of listening test data

Abstract

As there is a need for quantifying the perceived quality of automotive sounds in order to define target sound characteristics and set-up development targets, diverse jury test designs are frequently applied to collect data. The jury tests range from structured listening tests performed under controlled laboratory conditions to qualitative methods collecting free verbalization data during test drives. Though, jury tests are usually only used as an intermediate step to derive sound quality metrics based on psychoacoustic quantities allowing the quantitative prediction of subjective responses to product sounds. However, the quality of jury test data and the way of processing the data drive the validity of the developed sound quality metric. Therefore, a detailed quality check of listening test data is of particular importance but often neglected in practice. The quality check involves the investigation of intra- and inter-rater reliability issues, the detection of outliers and clusters, the use of adjusted error measures, the (cross) validation of sound quality metrics for estimating the generalizability of results. The different aspects in investigating listening test data in the context of vehicle noises and their suitability for sound quality metric development will be discussed on the basis of selected examples.