Toggle Main Menu Toggle Search

Open Access padlockePrints

Nonnegative matrix factorization 2D with the flexible beta-Divergence for Single Channel Source Separation

Lookup NU author(s): Dr Wai Lok Woo, Professor Satnam Dlay


Full text for this publication is not currently held within this repository. Alternative links are provided below where available.


This paper presents an algorithm for nonnegative matrix factorization 2D (NMF-2D) with the flexible beta-Divergence. The beta-Divergence is a group of cost functions parametrized by a single parameter beta. The Least Squares divergence, Kullback-Leibler divergence and the Itakura-Saito divergence are special cases (beta=2,1,0). This paper presents a more complete algorithm which uses a flexible range of beta, instead of be limited to just special cases. We describe a maximization-minimization (MM) algorithm lead to multiplicative updates. The proposed factorization decomposes an information-bearing matrix into two-dimensional convolution of factor matrices that represent the spectral dictionary and temporal codes with enhanced performance. The method is demonstrated on the separation of audio mixtures recorded from a single channel. Experimental tests and comparisons with other factorization methods have been conducted to verify the efficacy of the proposed method.

Publication metadata

Author(s): Yu KW, Woo WL, Dlay SS

Publication type: Conference Proceedings (inc. Abstract)

Publication status: Published

Conference Name: 2015 IEEE Workshop on Signal Processing Systems (SiPS)

Year of Conference: 2015

Print publication date: 01/01/2015

Acceptance date: 01/01/1900

Publisher: IEEE


DOI: 10.1109/SiPS.2015.7344990