Identifying salient sounds using dual-task experiments

Auditory saliency refers to the characteristics of a sound that cause it to attract the attention of a listener. Pre-attentive or bottom-up saliency has to do with automatic processing in the human auditory system that does not require and often precedes attention. Unlike visual saliency, where eye-tracking is a commonly used evaluation method, with auditory saliency, there is no easily trackable physical correlate that can be used for evaluation. Other auditory saliency models [1, 2] have been evaluated using tests that did not specifically target bottom-up saliency. In this paper, we present a method to conclusively isolate bottom-up auditory saliency. There are also several important applications to bottom-up saliency in auditory scene analysis, auditory display design and analysis, and speech processing.