Investigating Concurrent Auditory Icon Recognition

This paper an investigation of the identification of concurrently presented auditory icons. The motivation for this work was to get a better understanding of the identification of an everyday sound scene. We collected a set of descriptions for a set of everyday sounds as classified by the participants, using their free text responses. Two different experiments were conducted. The first experiment used no sub-categorisation or classification information when choosing the auditory icons. The second experiment used object and action descriptors in the selection of auditory icons. Our hypotheses was that by ensuring auditory icons did not have the same object or action descriptors, the identification of auditory icons would improve. Both experiments used an onset-to-onset gap of 300 ms between auditory icons. The results show that when there was no overlap between the object and the action descriptors of the concurrent auditory icons, the identification of the auditory icons was significantly improved.