基于频率和形状特征的脉冲重复间隔调制识别

Pulse Repetition Interval Modulation Recognition Based on Frequencies and Patterns

根据雷达信号脉冲序列的特点,从雷达脉冲信号中提取频率特征和形状特征,构成二维特征向量,并用支持向量机设计多类别分类器,实现雷达信号PRI调制信号的自动识别,实验结果表明,对特征向量进行大幅度降维(从64维降到2维)后,既简化了分类器,又保持或提高了识别率和抗噪声性能．与原特征向量相比,对无噪样本的误识率从0．15％-0．25％降低到0．00％,对有噪样本的误识率从0．40％-1．30％降低到0．15％- 0．93％．

According to the characteristics of pulse trains of radar signals, frequency and pattern are extracted from radar emitter signals. The two features constitute two-dimensional vectors, which are taken as inputs of a classifier designed by a support vector machine to identify the pulse repetition interval modulation of radar emitter signals automatically. Experimental results show that when the dimensions are lowered from 64 to 2, the extracted feature vector decreases the complexity of the classifier while maintaining or even enhancing the performances in recognition rate and noise suppression. Comparing to the original feature vector, the error rate of recognition of the extracted feature vector decreases from 0.15% - 0.25% to 0.00% for the samples without noises, and from 0.40% - 1.30% to0.15% - 0.93% for noised ones.