计算机与现代化

• 算法设计与分析 • 上一篇    下一篇

基于聚类的双说话人混合语音分离

  

  1. 广西大学计算机与电子信息学院,广西南宁530004
  • 收稿日期:2014-01-20 出版日期:2014-04-17 发布日期:2014-04-23
  • 作者简介:作者简介:吴春(1988),男,广西贺州人,广西大学计算机与电子信息学院硕士研究生,研究方向:计算机软件与理论。

Cochannel Speech Separation Based on Clustering

  1. Computer and Electronic Information College, Guangxi University, Nanning 530004, China
  • Received:2014-01-20 Online:2014-04-17 Published:2014-04-23

摘要:  

摘要: 针对许多基于训练模型的计算机听觉场景分析系统,在解决双说话人混合语音信号分离时需要依赖样本训练的有效性以及说话人的先验知识,提出一种基于聚类的单声道混合语音分离系统。系统先利用多基音跟踪算法对语音信号进行分析并产生同时流,然后通过最大化类内散布矩阵与类间散布矩阵的迹,搜索同时流的最佳分类,最终完成对双说话人的语音分离。该系统不需要训练语音模型,并且有效地改善了在双说话人混合语音信号的分离效果,为双说话人的语音分离提供了新的思路。

关键词:  , 计算机听觉场景分析, 双说话人语音分离, 聚类

Abstract:  

Abstract:  This paper proposes an unsupervised clustering approach for cochannel speech separation to solve the problem that many auditory scene analysis (CASA) systems using training model to require the availability of pretrained speaker models and prior knowledge of participating speakers. The system produces simultaneous streams of mixture signal through multipitch tracking algorithm, and searches for the optimal assignment of simultaneous speech streams by maximizing the between and withincluster scatter matrix ratio to separate the mixtures. The system does not require trained speaker models, improves obviously the performance of cochannel separation, which offers a good solution to separate cochannel speech.

Key words:

中图分类号: