本文应用广义似然比(GLR,Geaeralized Likelihood Ratio)分段技术于水声回波信号分析,得到很好的结果。显示了广义似然比分段技术处理在噪声淹盖下信号的能力。本文示出两种实际信号的分析结果,岛屿和浸没物体回波。这些信号都在强噪声淹没下,有一些信号入耳根本检测不到,而这种分段技术却检测出其边界和确定了边界的位置。这结果预示线性预测技术应用于水声信号处理是有前途的。 In this paper, the Geeralized Likelihood Ratio (GLR) segmentation technique is applied to the acoustic echo signal analysis, and good results are obtained. The ability of generalized likelihood ratio segmentation techniques to handle signals under noise flooding is shown. This article shows the results of two real-world signals, the island and the submerged object echo. These signals are submerged in strong noise, some of the signal is not detected in the ear, and this segmentation technology has detected its boundaries and to determine the location of the border. This result indicates that linear prediction techniques are promising for underwater acoustic signal processing.