A Classifier-Based Approach to Score-Guided Source Separation of Musical Audio

摘要

Audio source separation seeks to decompose an audio recording into several different layers corresponding to independent sources, such as different speakers, or, in our case, musical parts. Source separation is a formidable task; although the problem has received considerable attention in recent years, it is safe to say that it remains open.