Challenging ITU-T P.835: Searching for the optimal order of scales for assessing the quality of complex speech signals

Abstract

When evaluating the quality of speech samples which contain background noise, the International Telecommunication Unit (ITU) has released the ITU-T P.835 standard for assessing the quality on several scales. The scales contain ratings on signal quality, background noise annoyance, and the overall quality. Hereby, the order is fixed in the recommendation, such that the overall quality should be always assessed last, and the signal quality and background noise randomly either first or second. However, no clear argumentation is given why only the first two items are supposed to be randomized but the overall quality should be assessed last. Furthermore, research has shown that the order of scales has an effect on the judgment made. In this paper, we will show that the order of scales has an effect on the rating scores in this scenario. Therefore, we conducted an experiment conform to the recommendation, and altered the order of scales. We used a complex speech sample which contained different types and intensities of background noise; these were also suppressed using noise suppression algorithms. These may have an influence on the speech signal itself. We assessed the proposed scales and differed the order of those in a within subject study design.

BibTeX

@inproceedings{voigtantons2015oc2,
  author    = {Arndt, S. and Möller, S. and Antons, J.-N.},
  title     = {Challenging ITU-T P.835: Searching for the optimal order of scales for assessing the quality of complex speech signals},
  year      = {2015},
  booktitle = {Paper presented at the 41st annual congress of the Deutsche Gesellschaft für Akustik (DAGA 2015), Nürnberg, Germany},
  url       = {https://pub.dega-akustik.de/DAGA_2015/data/articles/000127.pdf},
}