大气低频噪声混合模型的MCMC参数估计

Parameter estimation for mixture model of atmospheric noise through MCMC method

大气噪声是低频通信中的主要干扰,且具有严重非高斯分布特性,对非高斯噪声模型的参数估计对于提高低频接收机的性能具有重要意义。设计了估计非高斯混合模型参数的马尔可夫链蒙特卡罗(Markov chainMonte Carlo,MCMC)算法,该算法通过构建贝叶斯层次模型,利用Gibbs抽样和M-H抽样更新迭代参数。利用乘积特性,将稳定分布作为等价的高斯分布来处理,并在层次模型中设置多个额外参数,以增强其灵活性。仿真实验与实测数据表明,该算法迭代收敛快、精度高,有很高的实用价值。

Atmospheric noise is the main interference in a low-frequency communication system, which is highly impulsive. So the work for estimating the parameters of model of non-Gaussian noises is of great significance to improve the performance of the low-frequency receiver. This paper proposes a Markov chain Monte Carlo （MCMC） method to estimate the parameters of a mixture model. The method updates the parameters through a Gibbs sampler and M-H algorithm, which are based on the Bayesian hierarchical model. The a stable distribution in the mixture model is equivalent to the normal distribution by using the product properties. An extra layer is added to the hierarchy for full flexibility. The result shows that the new method has a good performance, high precision and can be excellently applied in practice.