摘要
该文针对平坦衰落信道下存在信道参数差异的多天线接收信号联合参数估计和符号检测问题,提出一种基于变分贝叶斯的联合处理算法。算法直接利用多个接收数据流进行信息符号的估计,抑制传统信号合成与解调解耦处理带来的性能损失。将问题建模为已知多组观测数据条件下发送符号、信道传输时延、信道增益和噪声功率的联合最大后验估计问题。基于变分贝叶斯理论对该最大后验进行近似求解,在相对熵最小化的准则下,推导得到了各个待估参数解析形式的近似后验分布——变分分布。所提算法无需计算各参数精确的点估计值,而是采用信道参数和信息符号变分分布迭代处理的方式进行联合求解。仿真结果表明,所提算法通过多信号、多参数的联合处理能够获得优于经典解耦处理和部分联合处理技术的系统误码率性能,且在接收天线数目较多和观测数据长度较短时性能优势体现更加明显。
For the issue of joint parameter estimation and symbol detection for multi-antenna signals with channel parameters difference over fiat-fading channels, a new joint processing scheme is proposed based on the Variational Bayes (VB) method. The proposed scheme uses directly multiple received signals for the estimation of information symbols, restraining the information loss in conventional decoupled scheme of signals combination and demodulation. The problem is modeled as the joint Maximum A Posteriori (MAP) estimation of information symbols, time-delays, complex channel gains, and noise powers, given multiple observations, and approximately solved by means of VB approach. Based on the criterion of minimum relative entropy, analytical-form of the approximate distributions, i.e., variational distributions, for all unknown parameters are derived. There is no need to determine accurate point estimates of the parameters. Instead, the proposed scheme proceeds iteratively by alternating between the variational distributions of channel parameters and the information symbols. Simulation results show that the proposed joint processing scheme has significant performance improvements in comparison with conventional decoupled or partly joint processing schemes especially with large array sizes and short signal lengths.
作者
张凯
田瑶
谢云鹏
刘翼
ZHANG Kai;TIAN Yao;XIE Yunpeng;LIU Yi(Luoyang Electronic Equipment Test Center of China,Luoyang 471000,China;96862 Troops,Luoyang 471000,China)
出处
《电子与信息学报》
EI
CSCD
北大核心
2018年第9期2096-2104,共9页
Journal of Electronics & Information Technology
基金
国家自然科学基金(61501517)~~
关键词
多天线组阵
平坦衰落
联合处理
变分贝叶斯
Multi-antenna arraying
Flat-fading
Joint processing
Variational Bayes