重叠语音的帧同步分离研究  

A Study on the Frame Synchronized Segregation of Overlapped Speech

作　　者：戴礼荣[1] 宋彦[1] 王仁华[1] 

机构地区：[1]中国科学技术大学电子工程与信息科学系,合肥230027

出　　处：《电子学报》2002年第10期1552-1554,共3页Acta Electronica Sinica

摘　　要：重叠语音在本文中定义为由两个及两个以上原始语音叠加所形成的混合语音 .重叠语音分离 ,由于不能利用方位信息 ,困难更大 .但人的听觉系统却能够有效感知和分离这种重叠语音 .本文利用语音的谐波特性对重叠语音分离进行研究 .提出基于语音信号谐波分析的帧同步分离方法 ,并给出实验结果 .In this article,the overlapped speech is defined as the hybrid of two or more original speeches.Because there is not any direction information,it is more difficult to segregate the overlapped speech.But the human auditory system can successfully perceive and segregate this kind of overlapped speech.This paper addresses the problem of the overlapped speech segregation based on the speech harmonicity.A frame synchronized segregation method based on harmonic analysis is proposed and some corresponding experimental results are given.

关 键 词：重叠语音 帧同步分离 语音分离 基音频率 谐波分析 

分 类 号：TN912.3[电子电信—通信与信息系统]