Evaluating the use of crowdsourced data classifications in an investigation of the steelpan drum

Abstract

The effectiveness and reliability of crowd-sourced data classification to study the acoustics of the steelpan was evaluated. A project was developed and hosted on the widely used Zooniverse website. Volunteers on the project’s site were asked to identify areas of maximum vibrations (called antinodes) and number of bright rings (fringes) in those areas for each classification. We explored various methods in ensuring volunteers generate successful classifications. The data for classification comes from a high-speed video recording, paired with Electronic Speckle Pattern Interferometry, of a strike on the steelpan’s surface, which produces thousands of frames to be analyzed. We developed the project in preparation for a public release. We have analyzed the collected classifications using imported Python libraries. After validation and averaging of volunteer classifications, an Amplitude vs. Time graph was obtained for each contributing note in the recording of a strike.