Conference paper Open Access
Jordan Smith; Yuta Kawasaki; Masataka Goto
To create their art, remix artists would like to have segmented stem tracks at their disposal; that is, isolated instances of the loops and sounds that the original composer used to create a track. We present Unmixer, a web service that will analyze and extract loops from any audio uploaded by a user. The loops are presented in an interface that allows users to immediately remix the loops; if users upload multiple tracks, they can easily create mash-ups with the loops, which are automatically matched in tempo. To analyze the audio, we use a recently-proposed method of source separation based on the nonnegative Tucker decomposition of the spectrum. To reduce interference among the extracted loops, we propose an extra factorization step with a sparseness constraint and demonstrate that it improves the source separation result. We also propose a method for selecting the best instances of the extracted loops and demonstrate its effectiveness in an evaluation. Both of these improvements are incorporated into the backend of the interface. Finally, we discuss the feedback collected in a set of user evaluations.