Investigating timbral attribute perception using Best-Worst Scaling

Title: Investigating timbral attribute perception using Best-Worst Scaling
Authors: Victor Rosi
Source URL: https://www.timbreandorchestration.org/tor/modules/video-series/symposiums-workshops/y3/rosi

* if you can’t see the video because the music in the presentation is copyright protected in your country, contact the author at joshua.rosner[at]mail.mcgill.ca

Abstract:

There are several key attributes designated by sound experts to characterize timbre. Their use is common but varies greatly depending on the professional field. It is therefore not always guaranteed that their understanding is similar. In this work we want to better understand the perception of four subjective attributes of timbre and try to propose an acoustic definition. By using a subjective annotation method called Best-Worst Scaling on a large sound database, our results show differences in understanding of these attributes within and between four professional groups (i.e. sound engineers, conductors, sound designers and composers). Basically, Best-worst scaling allows to associate latent scores to each item of a corpus according to a studied subjective concept. The exploitation of these scores, coupled with the extraction of audio features from the sound database, will allow the emergence of the acoustic code of each of the studied attributes in the framework of a feature selection process. Our results will inform us about the relationship that may exist between the terms. The methodology is also presented as a potential solution to the annotation and processing of a large number of sound stimuli in an experimental setting.