Skip to content

Latest commit

 

History

History
16 lines (11 loc) · 1.05 KB

README.md

File metadata and controls

16 lines (11 loc) · 1.05 KB

Audio Source Separation Using Non-Negative Matrix Factorization (NMF)

The task of single channel audio source separation, which consists in isolating the different audio sources present in a mixture signal. This can be used in several applications from distinguishing the different music instruments in an orchestra recording to speech enhancement by separating relevant information of a speaker and the background noise.\

NMF techniques implemented from: Cédric Févotte, Emmanuel Vincent, Alexey Ozerov. Single-channel audio source separation with NMF: divergences, constraints and algorithms. Audio Source Separation, Springer, 2018. ffhal-01631185f https://inria.hal.science/hal-01631185/document

CNN model implemented from: Blaszke M, Kostek B. Musical Instrument Identification Using Deep Learning Approach. Sensors (Basel). 2022 Apr 15;22(8):3033. doi: 10.3390/s22083033. PMID: 35459018; PMCID: PMC9025072.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9025072/

Audio samples for NMF application were created in GarageBand by Kevin Lu using an M-Audio KeyRig49