| 引用本文: | 冯士民,周穗华,应文威.非高斯码元检测的马尔可夫链蒙特卡洛算法.[J].国防科技大学学报,2015,37(4):111-116.[点击复制] |
| FENG Shimin,ZHOU Suihua,YING Wenwei.Symbol detection algorithm in non-Gaussian noise using Markov chain Monte Carlo method[J].Journal of National University of Defense Technology,2015,37(4):111-116[点击复制] |
|
| |
|
|
| 本文已被：浏览 8416次 下载 5701次 |
| 非高斯码元检测的马尔可夫链蒙特卡洛算法 |
| 冯士民1, 周穗华1, 应文威2 |
|
(1. 海军工程大学 兵器工程系, 湖北 武汉 430033;2. 中国人民解放军91635部队, 北京 102249)
|
| 摘要: |
| 针对实际超低频接收机不仅受非高斯噪声的影响，还受接收机内部和外部环境中高斯噪声影响的问题，对噪声采用非高斯分布和高斯分布的混合模型建模，根据混合模型的性质，设计了一种利用马尔可夫链蒙特卡洛方法的超低频信号码元盲检测算法。盲检测算法在贝叶斯层次模型下，采用Gibbs抽样和M-H抽样更新参数，同步估计信道衰落系数和噪声模型参数，并实现对信号码元的检测。算法迭代效率快、精度高。通过与最优检测算法性能比较，盲检测算法性能优异，对超低频信号接收具有重要的现实意义。 |
| 关键词: 非高斯噪声 盲检测 高斯尺度混合 混合模型 |
| DOI：10.11887/j.cn.201504019 |
| 投稿日期：2014-08-29 |
| 基金项目:国家自然科学基金资助项目（51109215） |
|
| Symbol detection algorithm in non-Gaussian noise using Markov chain Monte Carlo method |
| FENG Shimin1, ZHOU Suihua1, YING Wenwei2 |
|
(1. Department of Weaponry Engineering, Naval University of Engineering, Wuhan 430033, China;2. The PLA Unit of 91635, Beijing 102249, China)
|
| Abstract: |
| Considering that the receiver was not only affected by the non-Gaussian noise but also affected by its internal and external environment of Gaussian noise, a mixed model composed by non-Gaussian distribution plus Gaussian distribution was proposed. A blind detection algorithm based on Markov Chain Monte Carlo method was designed according to the properties of the mixed model. The blind detection algorithm could estimate the channel fading coefficient, parameters of noise model and could detect signal element. Detect signals based on Bayesian hierarchical model was using Gibbs sample and M-H sample for parameter updating. The algorithm has a high iterative efficiency and precision. Results show that the proposed blind detection algorithm performs as well as the optimal detection algorithm and has important realistic significance in super low-freguency signal reception. |
| Keywords: non-Gaussian noise blind detection Gaussian scale mixture mixed model |
|
|
