The EnviSounds dataset aims to remedy the shortcomings of the existing resources. It comprises a comprehensive selection of 530 environmental sounds covering categories at different levels of complexity and different sounds produced by the same sound source. Complemented by the additional insight about sounds’ acoustic variability and norms from behavioural studies, it enables researchers to select items for experiments based on the desired categories or specific criteria across a range of acoustic features. The dataset can be used for specifying training sets with individual sound sources, defining the boundaries of various categories or providing additional variables to building prediction models. Our dataset allows for exploring long-neglected issues of within-class variability and between-class confusion by providing multiple exemplars of each class and a selection of acoustic descriptors and normative data. We believe it will be a valuable resource for behavioural and neuroimaging research.
