机构地区: 华南理工大学电子与信息学院
出 处: 《数据采集与处理》 2003年第4期389-392,共4页
摘 要: 考虑当用户序列存在时间相关性时的多用户检测 ,并假设这种相关性可以用 Markov链描述 ,在传统的线性最大似然检测器中嵌入一个隐 Markov模型估计过程。因为输入序列是 Markov链 ,检测器的输出可以看成是被噪声污染的 Markov序列 ,Markov模型估计子用于估计用户序列及其转移概率 ,而估计得到的用户序列用来更新检测器的估计。因此 ,检测器和用户序列可以通过迭代的方式求解。仿真结果显示本文算法能充分利用信道输入的时间相关性 。 The user inputs of a multi user system are time correlated and can be modeled as first order finite state Markov chains. A linear maximum likelihood (ML) detector with an embedded hidden Markov model (HMM) estimator is proposed. Since the inputs are Markov chains, the detector output can be considered as a Markov chains corrupted by noise, a Markov estimation procedure is employed to estimate the desired user sequence and its transition matrix. The estimated user symbols in turn are used to update the detector estimation. Therefore, the detector and the user symbols are iteratively estimated. Simulational results show the effectiveness of the algorithm.