摘要
语音分离是实现机器听觉的一个重要而基础性的任务,单通道语音分离是语音分离中最为困难的问题。讨论了基于模型的单通道语音分离方法,对说话人依赖的、说话人选择的和说话人独立的三类单通道语音分离问题展开分析,并指出当前方法存在的问题和影响算法性能的关键因素。最后对基于模型的单通道语音分离研究发展方向进行了展望。
Speech separation is an important and basic task for computer audition,and single-channel speech separation is the most difficult problem in the field of speech separation.This paper reviewed the research topic of model-based single channel speech separation and discussed major methods developed in the literature which could group into three major categories：speaker-dependent,speaker-selection and speaker-independent separation.Analyzed the major shortcomings of each method and studied the key factors that affected the performance of the algorithm.Finally,discussed the future research directions on this topic.
出处
《计算机应用研究》
CSCD
北大核心
2010年第11期4025-4031,共7页
Application Research of Computers
关键词
单通道语音分离
基于模型
说话人依赖
说话人选择
说话人独立
single-channel speech separation
model-based
speaker-dependent
speaker-selection
speaker-independent
