Using dynamic time warping to improve the classical music production workflow

Abstract

The current music production workflow, comprising recording, editing, mixing, and mastering music, requires a great deal of manual work for the sound engineer. This thesis aims to bring some recent advances in Music Information Retrieval (MIR) techniques to music production tools, with the goal of streamlining the current process followed by sound engineers. We explored all areas in the music production workflow (with a focus on classical music) that could benet from digital signal processing (DSP) and MIR-based tools, built and iterated on these tools, and transformed the tools into products that are beneficial and easy to use. We collaborated with the Boston Symphony Orchestra (BSO) sound engineers to gather requirements for this work, which led to the identification of our two tools: an automatic marking transfer (AMT) system and an audio search (AS) system. We then collaborated with other potential users for both AMT and AS tools, including sound engineers from radio stations in the Boston area. This enabled us to identify additional workflows and finalize requirements for these tools. Based on these, we created successful standalone applications for AMT and AS.