利用支持向量机提高水声信号的检测能力  

Improving underwater signal detection capability by using support vector machine

作　　者：吴克桐[1,2] 王永丰[1,2] 李磊[1,2] 蔡惠智[1] 

机构地区：[1]中国科学院声学研究所,北京100190 [2]中国科学院研究生院,北京100190

出　　处：《声学技术》2009年第3期222-226,共5页Technical Acoustics

摘　　要：混响是影响主动声纳检测的一个重要因素。传统方法如高阶统计方法和修正协方差等白化方法在检测静止目标时效果并不理想,主要是因为这些算法无法利用混响这样的统计多变且非线性的数据准确估计一个AR模型的参数。利用支持向量机SVM(Support Vector Machine,SVM)来估计AR模型的参数,并进行了实验。结果表明,即使在静止目标的情况下,该算法依然能表现出良好的性能。Reverberation is one of the most important factors that degrade the performance of active sonar detection. Traditional methods such as HOS prewhitening and modified-covariance prewhitening method have difficulties in dealing with static targets. One of the most dominant reasons is that such methods have poor ability to estimate parameters of an AR model from the nonlinear and varying statistics reverberation data. A novel method using support vector machine （SVM） to estimate the parameters of an AR model is presented. Experimental results show that the approach improves the detection performance significantly even if the target is static.

关 键 词：支持向量机 混响 信号检测 

分 类 号：TN911[电子电信—通信与信息系统]