一种新的贝叶斯调制分类算法

A Novel Bayesian Modulation Classification Algorithm

提出了一种基于马尔可夫链蒙特卡罗(MCMC)的数字调制分类方法。针对存在未知残留载波相位和频率时贝叶斯分类难以实现的问题,采用Metropolis-Hastings(M-H)算法估计边缘似然概率密度,从而在分类性能上保持了贝叶斯分类的理论最优性和稳健性。利用对比实验验证了方法的性能。

A novel method is proposed for digital modulation classification based on Markov chain Monte Carlo （MCMC） Considering the difficulty for Bayesian classifier with unknown residual carrier phase and frequency, marginal likelihood probability density is estimated by Metropolis-Hastings （M-H） algorithm, which kept the theoretical optimality and robustness of Bayesian classifier, The simulated results show that the novel classifier outperforms the one based on cumulants.