中文摘要：

在研究LMS自适应算法的基础上,提出一种基于声门脉冲的变步长LMS自适应时延估计新方法,并在相关噪声和混响的环境下与互功率谱相位广义互相关法（GCC-CSP）、变步长LMS自适应算法进行性能比较。实验结果表明,新方法具有很好的鲁棒性,即使在低信噪比强混响的环境下也能获得有效的时延估计。

英文摘要：

This paper presented a speech signal variable step size LMS time delay estimation method based on glottal pulses in detail. Considered with the presence of correlated noise and reverberation, assessed the validity of the new algorithm by experiments with fan noise source in simulated reverberant room. Furthermore,compared it with the GCC-CSP method and the variable step size LMS algorithm. As a result that the new approach is superior and robust. Even in the environment of low SNR and strong reverberation, the proposed method can also give effectual time estimation.