混合语音信号的盲源分离技术研究  

Research on Blind Source Separation of Mixed Speech Signals

作　　者：邹彤 王英[1] 董姝敏 隋鹏 Zou Tong;Wang Ying;Dong Shumin;Sui Peng(School of Electronic Information,Qingdao University,Qingdao 266071,China;Dalian Maritime University)

机构地区：[1]青岛大学电子信息学院,青岛266071 [2]大连海事大学

出　　处：《单片机与嵌入式系统应用》2019年第8期14-17,21,共5页Microcontrollers & Embedded Systems

摘　　要：本文主要研究了常见的三种混合模型(线性瞬时混合、线性卷积混合和非线性混合)的盲源分离问题,阐述了它们各自的原理和方法,并讨论了算法好坏的衡量方法和优化方法。本文对众多学者提出的改进的仿生智能优化算法进行了整理与总结。最后,对待解决的语音信号盲源分离问题进行了归纳总结,为进一步研究提供了明确的方向。In the paper,three common types of blind source separation are studied:linear instantaneous mixing,linear convolutional mixing and nonlinear mixing.Their principles and methods are summarized,and the methods to measure and optimize the algorithm are discussed.In this paper,the improved bionic intelligent optimization algorithm proposed by many scholars is sorted out and summarized.Finally,the problem of blind speech separation is summarized,which provides a clear direction for further research.

关 键 词：语音信号 盲源分离 非平稳信号 最大似然估计 仿生智能优化算法 

分 类 号：TN912.3[电子电信—通信与信息系统]