Cross-synthesis using non-negative Matrix Factorisation (NMF)
NMFMorph uses a technique called Optimal Transport to provide real time interpolation between
target bases of a BufNMF decomposition. Using Optimal Transport for the spectral interpolation provides richer results than a simple linear interpolation between spectral shapes. Activations from a BufNMF analysis are also provided to activate the interpolated spectral information through time.
autoassign is set to 1, NMFMorph will determine which bases from
target best match each other, and will interpolate between the matched pairings. If
autoassign is set to 0, NMFMorph will interpolate between bases in the order they have been provided.
Audio Morphing using Matrix Decomposition and Optimal Transport
Paper by Roma, Green, & Tremblay describing the NMFMorph algorithm. (begins on PDF page 165 of the DAFx2020 Proceedings)https://pure.hud.ac.uk/en/publications/audio-morphing-using-matrix-decomposition-and-optimal-transport
Audio Transport: A Generalized Portamento via Optimal Transport
The paper which inspired this implementationhttps://arxiv.org/pdf/1906.06763.pdf
Computational Optimal Transport
For those who really want to get down in the detailshttps://arxiv.org/pdf/1803.00567.pdf
Optimal transport: a hidden gem that empowers today’s machine learning
A relatively easy to approach article on applications of optimal transporthttps://towardsdatascience.com/optimal-transport-a-hidden-gem-that-empowers-todays-machine-learning-2609bbf67e59