摘要
论文研究了支持向量机核函数构成条件以及不同核函数的特性,结合水中目标识别技术特点,提出了一种组合核函数支持向量机的方法。提取了基于小波变换的舰船辐射噪声奇异性、尺度-过零、尺度-能量特征,对水中目标进行了SVM分类识别。研究表明,基于组合核函数的支持向量机分类识别效果优于单独核函数的支持向量机识别效果。
The formation conditions and the characteristics of different kernel functions are studied. Based on the research of underwater target recognition, a method for training support vector machine with a hybrid kernel is proposed. After extracting characteristics of an underwater target such as Lipschitz singularity, scale zero-cross density and scale energy factor by using wavelet transformation, the hybrid kernel function SVM is used to recognize different targets. Experimental results indicate that using the hybrid kernel can give better performance compared to a single common kernel.
出处
《声学技术》
EI
CSCD
北大核心
2005年第3期144-147,共4页
Technical Acoustics
关键词
支持向量机
核函数
目标识别
support vector machine
kernel function
target recognition
