摘要： 传统的支持向量机(SVM)多值分类算法在构造多个二值分类器时存在计算量较大和分类速度较低的问题。为此，将覆盖思想引入SVM分类算法中，提出一种基于覆盖的SVM多分类算法，通过构造覆盖集寻找更紧致的优化区域，从而提高分类速度。将该算法应用到雷达辐射源识别中，仿真结果表明，该算法能够获得较好的识别效果。

关键词: 覆盖算法, 支持向量机, 超球面, 核函数, 辐射源识别

Abstract: Problem of big computation volume and low classifying velocity exists in conventional Support Vector Machine(SVM) multi-value classifying algorithm in constructing several two-value classifiers. For this, a SVM multi-value classifying algorithm based on covering is put forward. The covering thought is introduced into SVM algorithm and the tighter optimizing area can be found by constructing covering, which makes the classifying velocity improve notably. The algorithm is applied in radar emitter recognition. Simulation results show it is of excellent performance.

Key words: covering algorithm, Support Vector Machine(SVM), hypersphere, kernel function, emitter recognition

中图分类号: 

• TP311.52