Machine Audition Curriculum and Real-Time Music Accompaniment

Abstract

A machine audition curriculum was created as part of the MIT Media Lab’s Artificial Intelligence Education initiative. This curriculum was geared towards middle school students to help them understand how humans and machines perceive sound, and allow them to apply this knowledge to create and analyze their own music. This thesis presents the tools created to aid in the teaching of this curriculum: a new music audition Scratch extension. This extension introduces the ability to create and analyze music, as well as the integration of Google Magenta, a machine learning library that allows students to generate new music or accompany music that they have created. Through the use of this Scratch extension, it was possible to pilot the machine audition curriculum with middle school students and show that they were able to better understand signal properties, create and analyze their own music, and understand the similarities and differences between human and machine audition.