基于Fano译码复杂度和隐马尔科夫模型的信道建模和预测

【作者】 王春；

【导师】 孙锦涛； 韩玉兵；

【作者基本信息】 南京理工大学 ， 通信与信息系统， 2008， 硕士

【摘要】 无线通信信道的模型研究是无线通信领域的关键技术之一。相对于波形信道,离散信道模型(DCM,Discrete Channel Model)一般采用马尔科夫链(MarkovChains)进行描述,具有更高的计算效率,由于它把信道划分成不同的离散状态,在不同的状态下可以选择不同的自适应调制编码技术,这也是目前无线通信发展的趋势之一。在信道模型建立的条件下,准确的信道预测是信道自适应的另一个关键问题。本文从三个方面进行深入研究:(1)分别研究了信道信噪比和瑞利衰落信道多普勒频扩与Fano译码复杂度之间的对应关系。从统计意义上说,低信噪比条件下的Fano译码复杂度要大于高信噪比条件下的Fano译码复杂度,高多普勒频扩条件下的Fano译码复杂度要大于低多普勒频扩条件下的Fano译码复杂度,所以用Fano译码复杂度衡量信道条件是可行的。(2)研究了基于Fano译码复杂度和隐马尔科夫模型的离散信道建模。本文以量化的Fano译码的复杂度作为观测值,采用HMM中的Baum-Welch算法训练信道参数,并且在得到训练参数后用Viterbi译码算法来估计离散信道模型状态。最后采用随机松弛(SR,Stochastic Relaxation)优化算法改进传统的Baum-Welch算法,提高了离散信道模型参数估计精度,信道状态估计的准确率达到90%以上。(3)在离散信道隐马尔科夫模型建立的基础上,研究基于马尔科夫链转移概率矩阵的信道状态预测。首先通过滑动窗方法(Sliding Window)截取一定长度的观测值序列进行Viterbi译码得到当前时刻的信道状态;然后采用隐马尔科夫模型的状态转移概率矩阵进行信道状态的单步和多步预测,其中一步预测的效果非常好,多步预测因为误差累计的原因效果逐渐下降。更多还原

【Abstract】 The research on the model of wireless channel is one of the key technologies in the field of wireless communications. Compared to wave channels, discrete channel models usually adopt Markov model, which has higher computational efficiency. And because the Markov model divides the channel into different discrete states, we can choose different adaptive technology, which is also the trend in the wireless communication development. Another core problem of adaptive technology is the accurate prediction of channels.This thesis focuses on three aspects:Firstly, the relationship between the complexity of Fano sequential decoding, signal-to-noise ratio (SNR) in a mobile channel and Doppler spread in Rayleigh fading channels is disscussed. We found that when SNR is higher, the complexity of Fano sequential decoding is lower, and when Doppler spread is lower, the complexity of Fano sequential decoding is lower. So it is reasonable to use the complexity of Fano sequential decoding as a measure of the condition of channels.Secondly, Model-building of the discrete channel, using the complexity of Fano sequential decoding and HMM, is discussed. In this thesis, measured Fano decoding complexity is used as observation value, Baum-Welch algorithm is employed to obtain an optimal estimate of the HMM model parameters, and then the maximum likelihood hidden state sequence is found by using the Viterbi algorithm. After optimized Baum-Welch algorithm by Stochastic Relaxation algorithm, accuracy of the Discrete Channel Model estimate is improved and estimates accuracy rate of the channel state is over of 90 percent.Thirdly, research on the channel prediction with the state transition probability matrix of Markov model, based on discrete channel models by HMM is discussed. First, by using Sliding Window algorithm to obtain a sequence of measured values before performing Viterbi decoding, the instantaneous state of channel is achieved; Then using HMM’s state transition probability matrix to predict one step and many steps of channel states. One step of prediction has achieved good results; while, because of accumulated error, the result of prediction of many steps fails.更多还原