非负组合模型及其在声源分离中的应用  

Non-negative Compositional Models and Its Application in Acoustic Source Separation

作　　者：张雄伟[1] 李轶南[1] 时文华[1,2] 胡永刚[1] 陈栩杉[3] Zhang Xiongwei Li Yinan Shi Wenhua Hu Yonggang Chen Xushan(College of Command Information System, PLA University of Science and Technology, Nanjing, 210007, China Flight Instructor Training Base, Air Force Aviation University, Bengbu, 233000, China Lab of Political Information, People＇s Armed Police Institute of Politics, Shanghai, 201703, China)

机构地区：[1]解放军理工大学指挥信息系统学院,南京210007 [2]空军航空大学教官基地,蚌埠233000 [3]武警政治学院政工信息化教研室,上海201703

出　　处：《数据采集与处理》2017年第2期266-277,共12页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金(61471394;61402519)资助项目;江苏省自然科学基金(BK20140071;BK20140074)资助项目

摘　　要：非负组合模型在人工智能、数据挖掘和智能信息处理研究领域具有十分重要的应用意义,已经逐渐成为声源分离中最常使用以及最具代表性的模型之一。内含于其中的非负成分的加性组合与人类听觉系统的感知机理高度契合。利用非负组合模型进行声源分离的技术正在变得越来越流行。本文从被称作非负矩阵分解的最基本的非负组合模型开始,首先回顾了非负组合模型的基本原则,包括需要求解的基本问题、目标函数的度量以及求解相关问题的常用方法。在此基础上,系统地讨论了非负矩阵分解在声源分离不同应用领域的拓展。最后指出并讨论非负组合模型研究中有待进一步研究的开放问题。Non-negative compositional models are of great importance in the application ot artificial intelligence, data mining and intelligent information processing research. They have gradually become one of the most representative and frequently used models of acoustic source separation in recent years. The embedded additive combination of non-negative components matches well with the characteristic of human perception. Techniques that make use of non-negative compositional models have been increasingly popu- lar in acoustic source separation. Starting from the most basic non-negative compositional model, which is termed as non-negative matrix faetorization （NMF）, we firstly review the principles of non-negative some typical methods to solve related problems. Based on these principles, we systematically discuss the variety extensions of NMF designed for particular applications in acoustic source separation. Finally, some open problems are presented and discussed.

关 键 词：声源分离 非负组合模型 非负矩阵分解 

分 类 号：TN912.3[电子电信—通信与信息系统]