Effect of masker selection schemes on the perceived affective quality of soundscapes: A pilot study

Abstract

The perceived quality of a given soundscape may be improved by adding "pleasant" sounds via a speaker or electroacoustic system to the existing soundscape as maskers. Ideally, the choice of sounds should be automated to minimize the time and labor required to maintain such a system. Hence, we investigate the effect of several automated masker selection schemes on improving the ISO Pleasantness of urban soundscapes in a controlled laboratory environment. An ambisonic recording and 360-degree video of a location exposed to traffic noise was made and reproduced via a stacked planar speaker array and virtual reality headset. Simultaneously, maskers from a fixed candidate pool were overlaid on the reproduced soundscape in one of the following methods: random selection across the entire pool, fixed selection based on results in the literature, random selection across a restricted pool based on model-agnostic analysis of a pre-existing dataset, and random selection across a restricted pool based on a deep neural network trained on a pre-existing dataset. We also investigated a control condition with no added maskers, and evaluated the results based on changes in ISO Pleasantness ratings given by participants to the soundscape under each masker selection scheme, relative to the control condition.