Abstract: To improve the accuracy rate of radar signal recognition in increasingly sophisticated electronic countermeasure environment, an improved S3VM algorithm based on HSS（Heuristic Sampling Search） is put forward. According to theory of ambiguity function, through SVD of radar signal recognition’ ambiguity function, it extracts its singular value as characteristic parameters; aiming at disadvantages of traditional semi-supervised support vector machines, the classifier constructed by improved S3VM is used for classifying the radar signal, then the recognition of tested samples is finished. Improved S3VM try to exploit multiple representative large-margin low-density separators by using HSS. It works out the shortcomings on S3VM efficiently, such as low classification accuracy and unstable classification performance. The experimental results show that accuracy rate is improved obviously by the algorithm in radar signal recognition.

Key words: radar signal recognition, semi-supervised Support Vector Machine（SVM）, separators, ambiguity function, Singular Value Decomposition（SVD）

摘要： 为提升在日趋复杂的电子对抗环境中对雷达信号识别的准确率，提出了一种基于启发式采样搜索（Heuristic Sampling Search，HSS）改进S3VM的雷达辐射源信号识别算法。根据模糊函数理论，通过对雷达信号的模糊函数进行奇异值分解（SVD），提取出奇异向量作为雷达信号识别的特征参数；针对传统的半监督支持向量机（Semi-supervised SVM，S3VM）的不足，利用改进的S3VM构建分类器对雷达信号进行分类，完成对测试样本的识别。该方法通过启发式采样搜索来寻求具有代表性的多个大边缘低密度的分类决策面，有效解决传统S3VM分类精度低且分类性能不稳定等缺点。实验结果表明，在雷达信号识别中，该算法明显提高了分类准确率。

关键词: 雷达信号识别, 半监督支持向量机, 分类决策面, 模糊函数, 奇异值分解