中文摘要：

论文分析了基于语音激励信息的时间延迟算法分类，描述了基于EGCI的时延估计方法基本原理。研究了线性预测LPC残留误差时延估计法（LPCRE）提取时延的方法，并与另一种提取GCI最大似然估计法MLED和基于广义互相关的GCC—ML进行了性能分析和比较。结果表明基于EGCI的时延估计方法用于声源定住完全可行，其性能依赖于提取GCI的算法模型：三种算法中的LPCRE的均方偏差最小，抗混响能力强，运算量适中，便于实时实现。

英文摘要：

This paper analyzes the new type of method for localizing the source using the excitation information in speech,and presents the theory of time-delay estimation based on the EGCI.The time-delay estimation using the LPCRE is studied,which compared with MLED and G.CC-ML.The results shows that the method of using the excitation information in speech is totally used for the source localization depending on the model of EGCI, and the LPCRE is robust with the lest root of variance and the strong antl-reverberant, which is easy for real-time implement.