基于ReliefF和聚类的特征选择方法及其在无线电信号识别中的应用

Application of Feature Selection Approach Based on ReliefF and Clustering in Radio Signal Recognition

无线电信号识别在无线电监测中占有重要地位,为了提高信号识别率,针对C波段信号特征数据的特点,提出一种基于ReliefF和聚类的特征选择方法.该算法首先用ReliefF算法去除与分类不相关的特征,再对余下的特征根据相关度强弱进行特征聚类,最后根据特征权重大小和相关度强弱删除冗余特征,选出代表性特征.实验结果表明,该算法有效地减少了数据维数,并且提高了信号识别率.

Radio signal recognition is very important in the radio monitoring.According to the characteristics of C band signal feature data,a new feature selection approach based on ReliefF and clustering is presented in order to improve the signal recognition rate in this paper.Firstly this algorithm eliminates those features irrelevant to class making use of ReliefF algorithm,then clusters the rest features based on similarity measure,finally deletes redundant features and selects representative characteristics based on the weights and the similarity.Experiments show that the algorithm can effectively reduce the data dimension and improve the signal recognition rate.