关键词: 信号分选, 聚类算法, 支持向量聚类, 锥面聚类分配, 相似熵, 有效性验证, 分离性

Abstract:

The radar signal sorting method based on the traditional Support Vector Clustering(SVC) algorithm leads to a high time complexity, and the traditional validity index can not indicate the best sorting result efficiently. Aimed at the problem, a new sorting method is presented based on the Cone Cluster Labeling(CCL) method for the SVC algorithm. The CCL method relies on the theory of approximate coverings both in teature space and data space. And a new cluster validity index, Similitude Entropy (SE) index,is proposed which assesses the compactness and separation of clusters using the information entropy theory. The problem mentioned above can be resolved well with the proposed method. Experimental results show that the strategy can improve efficiency without sacrificing sorting accuracy. Meanwhile the proposed method can meet the real-time and accuracy requirements of the ELINT system to some extent and is of good practical value.

Key words: signal sorting, clustering algorithms, support vector clustering, cone cluster labeling, similitude entropy, verification, separation