摘要
目标识别是战场低空飞行目标声预警技术的核心内容之一。为了满足声预警系统的要求,建立的识别器必须高效、具有较好的推广能力。采用了一种新的分类器一支持向量机对目标进行了分类识别。首先简要描述了直升机、巡航导弹的声信号特性,说明了支持向量机的原理。以自回归模型参数为特征向量对3种直升机、一种巡航导弹共4类目标进行了识别,并同一种前向反馈神经网络进行了识别比较。支持向量机和该神经网络得到的识别率分别为88.1%和84.1%,结果表明此方法的有效性。最后分析了两种分类器识别错误的原因,给出了提高识别率的建议。
Passive acoustic recognition for helicopters and cruise missiles in battlefields has attracted great importance. A novel classifier-support vector machine (SVM) is adopted to identify 3 kinds of helicopters and one kind of cruise missile is used The acoustic signal characteristics of helicopters and cruise missiles are described briefly with an introduction to the principle of support vector machine. After the feature extractions based on autoregression model parameters, the support vector machine achieves a recognition rate of 88.1%. A back propagation (BP) neural network is also applied to identify the above 4 targets and the recognition rate is 84.1%. The cause for the error recognition and some advices for the future work are given in the end.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2005年第1期46-48,共3页
Systems Engineering and Electronics
基金
国防预研基金资助课题(43070101).
关键词
被动声预警
目标识别
支持向量机
神经网络
巡航导弹
Passive acousic early warning
Target recognition
SVM
neural network
Cruise missile
