Blind source separation by nonnegative matrix factorization with minimum-volume constraint
Blind source separation by nonnegative matrix factorization with minimum-volume constraint
复制标题
DOI：
10.1109/icicip.2010.5565228
复制
发表时间：
2010-09
期刊：
影响因子：
--
通讯作者：
Zuyuan Yang;Guoxu Zhou;Shuxue Ding;S. Xie
中科院分区：
文献类型：
--
作者：
Zuyuan Yang;Guoxu Zhou;Shuxue Ding;S. Xie
近年来，非负矩阵分解（NMF）因其广泛的应用前景而受到越来越多的关注。然而，仍然存在的一个问题是，由此产生的因素不一定是现实可解释的。通常在标准NMF中加入一些约束条件来生成这样的解释性结果。本文提出了一种基于自然梯度优化的最小体积约束NMF算法，并给出了一种有效的乘性更新算法。所提出的方法可以应用于盲源分离（BSS）问题，一个热门的话题，有许多潜在的应用，特别是如果源是相互依赖的。图像盲分离的仿真结果表明了该方法的优越性。
Recently, nonnegative matrix factorization (NMF) attracts more and more attentions for the promising of wide applications. A problem that still remains is that, however, the factors resulted from it may not necessarily be realistically interpretable. Some constraints are usually added to the standard NMF to generate such interpretive results. In this paper, a minimum-volume constrained NMF is proposed and an efficient multiplicative update algorithm is developed based on the natural gradient optimization. The proposed method can be applied to the blind source separation (BSS) problem, a hot topic with many potential applications, especially if the sources are mutually dependent. Simulation results of BSS for images show the superiority of the proposed method.