Source Separation using Non-negative Matrix Factorization

PLEASE USE EARPHONES. This video introduces source separation using non-negative matrix factorization (NMF). It covers some standard steps in source separation and more specifically, the use of NMF and its variants for the same. Some familiarity is assumed with concepts from linear algebra such as matrix products and positive definiteness, elementary calculus and Fourier transforms. A detailed report and vectorized code for all the simulations can be found on GitHub: https://milind-blaze.github.io/project/research/source-separation/ More details on the Frobenius Perron theorem can be found at the Wikipedia page: https://en.wikipedia.org/wiki/Perron%E2%80%93Frobenius_theorem The scripts that were used to make this video can be found on GitHub: https://tinyurl.com/yba5k238 These scripts are based on the project developed by the brilliant 3Blue1Brown: http://www.3blue1brown.com/about/ https://www.youtube.com/channel/UCYO_jab_esuFRV4b17AJtAw/featured This video was made as a part of the linear algebra course EE5120: Applied linear algebra for Electrical Engineers taught at IIT Madras. The central ideas come from the following papers: “Monaural sound source separation by nonnegative matrix factorization with temporal continuity and sparseness criteria.” - Virtanen, Tuomas. https://tinyurl.com/y8nr5s3t “Algorithms For Non-negative Matrix Factorization.” - Daniel D. Lee, H. Sebastian Seung. https://tinyurl.com/y8s44726 “The Why and How of Nonnegative Matrix Factorization.” - Nicolas Gillis. https://tinyurl.com/yah6333h