CochlScene: Acquisition of Acoustic Scene Data Using Crowdsourcing

Cochl Acoustic Scene Dataset, or CochlScene, is a new acoustic scene dataset whose recordings are fully collected from crowdsourcing participants. Most of the initial plans and guidelines for the processes were provided by the researchers in the field of audio signal processing and machine learning including the authors, and the actual process was performed by using the crowdsourcing platform developed by SelectStar, a Korean crowdsourcing data company. During the process, the initial plans were reinforced and modified from the discussion about the actual difficulty in the collection process. After extracting the subset of the total collections considering the purpose of the data, we collected 76,115 10 seconds files in 13 different acoustic scenes from 831 participants.