The Nonnegative Matrix Factorization (NMF) is widely used in audio source separation tasks. However, the separation quality of NMF varies a lot depending on the mixture. In this paper we analyze the use of NMF in source separation tasks and show how separation results can be significantly improved by using the Nonnegative Matrix Factor 2D Deconvolution (NMF2D). NMF2D was originally proposed as an extension to the NMF to circumvent the problem of grouping notes, but it is used differently in this paper to improve the separation quality, without taking the problem of grouping notes into account.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.