Toward an Intelligent Editor of Digital Audio: Recognition of Musical Constructs

perform operations and transformations whose realization by signal processing techniques alone would range from cumbersome to unimaginable. In the cumbersome category might be such operations as "advance 15 sec and stop at the first G" or "play from the loudest point near 25 sec to the loudest point near 50 sec." In the barely imaginable category might be "start at the 17th measure" or "replay the violin part, this time transposed down a fifth and with the timbre of a clarinet, while preserving all stylistic features of the original." This article and its companion, by Foster, Schloss, and Rockmore, address methods of analysis that make such operations imaginable and maybe even practical. Here, we will focus on higher-level approaches to the problem. The companion article, which follows this one, will focus on lower-level methods.