Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 260 - Statistics and AI in Music
Type: Topic Contributed
Date/Time: Tuesday, August 4, 2020 : 1:00 PM to 2:50 PM
Sponsor: Royal Statistical Society
Abstract #309653
Title: Understanding Audio from Music Practice Sessions
Author(s): Christopher Raphael*
Companies: Indiana University
Keywords: music; practice; dynamic programming; sonfication; visualization

In a common scenario for musical practice a player works from a musical score, repeating problematic sections, jumping around in the score, and practicing exercises derived from the score. We present ongoing work in using the audio of a series of such unscripted practice sessions to give useful feedback to a musician. Before one can make useful criticism one needs to know what the player intended to play, thus score alignment plays a key role. A simple dynamic programming approach works reasonably well at dividing the audio into excerpts and matching each excerpt to the score. Such descriptions enable illuminating visualizations, such as those showing the coverage of the piece, the tuning errors, and obvious rhythm problems. These latter measures can be shown on a given excerpt, on a "best performance" pieced together from excerpts, or averaged over the entire collection of excerpts. We will present a method of computing such a best performance that provides a measure of the player's overall mastery of the piece at hand. We also explore sonification experiments designed to depict the population of excerpts.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program