Using Interactive Machine Learning to Sonify Visually Impaired Dancers' Movement

Abstract

This preliminary research investigates the application of Interactive Machine Learning (IML) to sonify the movements of visually impaired dancers. Using custom wearable devices with localized sound, our observations demonstrate how sonification enables the communication of time-based information about movements such as phrase length and periodicity, and nuanced information such as magnitudes and accelerations. The work raises a number challenges regarding the application of IML to this domain. In particular we identify a need for ensuring even rates of change in regression models when performing sonification and a need for consideration of how to convey machine learning approaches to end users.