一种基于相关向量机的调制方式识别算法

Modulation Classification Based on Relevance Vector Machine

针对数字信号的调制方式识别问题,给出了一种基于相关向量机的分类方法。相关向量机基于贝叶斯学习方法,其判决函数仅取决于训练样本的一小部分。文章提取信号的谱相关特征参数,设计了合理的分类策略。实验结果表明,与支持向量机相比,基于相关向量机的分类方法在保持较高识别率的同时,提高了调制识别的时效性。

This paper addresses the problem of automatic modulation recognition of digital signals.A classification method based on relevance vector machine（RVM） is developed.RVM is based on Bayesian estimation theory,and as a feature it uses a linear combination of kernel functions centered on a very small number of the training data.The spectral correlation theory is introduced and several useful characteristic parameters are extracted.Experiment results show that the RVM classifier achieves essentially the same recognition performance as the SVM classifier,but greatly reduces the computational complexity.