论文部分内容阅读
依据星座图采用非参数贝叶斯方法对多元相移键控(MPSK)信号进行调制识别.将未知信噪比(SNR)水平的MPSK信号看成复平面内多个未知均值和方差的高斯分布依照一定的比例混合而成,利用非参数贝叶斯推断方法进行密度估计,实现对MPSK信号分类目的.推断过程中,引入Dirichlet过程作为混合比例因子的先验分布,结合正态逆Wishart(NIW)分布作为均值和方差的先验分布,根据接收信号,利用Gibbs采样的MCMC(Monte Carlo Markov chain)随机采样算法,不断调整混合比例因子、均值和方差.通过多次迭代,得到对调制信号的密度估计.仿真表明,在SNR>5dB,码元数目大于1600时,2/4/8PSK的识别率超过了95%.
Nonparametric Bayesian (MPSK) signal is modulated according to the constellation map, and MPSK signal with unknown signal-to-noise ratio (SNR) is considered as Gaussian distribution with unknown mean and variance in complex plane According to a certain proportion, we use the nonparametric Bayesian inference method to estimate the density and achieve the purpose of MPSK signal classification.In the process of inference, the Dirichlet process is introduced as a priori distribution of mixed scale factor, combined with normal inverse Wishart (NIW ) Distribution as a priori mean and variance distribution, according to the received signal, the use of Gibbs sampling MCMC (Monte Carlo Markov chain) random sampling algorithm, continue to adjust the mixing ratio factor, mean and variance.Through multiple iterations, the modulation signal Density Estimation Simulation shows that the 2/4 / 8PSK recognition rate exceeds 95% at SNR> 5dB with more than 1600 symbols.