认知无线电中基于Markov模型的频谱预测算法研究

发布时间：2018-04-01 14:29

本文选题：认知无线电　切入点：频谱预测　出处：《西安电子科技大学》2014年硕士论文

【摘要】：频谱预测是认知无线电中的一项关键技术,它是指认知系统根据获得的信道历史信息,分析频谱的使用规律,进而预测频谱的空洞信息,以指导认知设备进行智能的频谱感知和动态的频谱接入,从而降低主次用户的碰撞率,提高认知系统的整体性能,最终达到提高频谱利用率的目的。由于现有频谱预测算法复杂度高、准确度低、所需历史序列长,不适用于能量受限的认知节点和快变信道环境。本文以提高预测准确度降低预测复杂度为目的,提出一种基于上下文树的可变长Markov预测方法。该方法将认知用户感知的历史状态信息构建成可变阶的状态树,在进行预测时动态改变所需的最近历史状态数目,具有复杂度低、预测准确度高、所需历史序列短的特点,并能通过周期的更新状态树提高非平稳环境下的预测准确度。考虑到频谱检测误差对预测准确度的影响,进一步完善上述算法,提出一种隐马尔科夫模型与上下文树可变长Markov模型相结合的频谱预测方法,先是训练隐马尔科夫模型恢复真实的信道状态序列,然后在此基础上构建状态树,从而取消频谱检测误差对预测性能的影响。利用基于排队模型产生的频谱数据,在平稳环境和非平稳环境下分别验证了基于上下文树可变长Markov方法的有效性。又分别利用排队模型和离散时间Markov模型产生的频谱使用数据,验证了在检测误差存在的情况下,隐马尔科夫模型与上下文树可变长Markov模型相结合的算法的有效性。
[Abstract]:Spectrum prediction is a key technology in cognitive radio, it refers to the cognitive system according to the channel history information, analysis of use of spectrum, then forecast the cavity spectrum, spectrum sensing and dynamic spectrum access to intelligence to guide the cognitive devices, thereby reducing the collision rate of the primary and secondary users, improve the overall performance of cognitive the system, finally achieve the purpose of improving spectrum efficiency. Because the existing spectrum prediction algorithm with high complexity and low accuracy, the history of long sequences, cognitive nodes do not apply to the limited energy and fast varying channel environment. In order to improve the prediction accuracy for the purpose of reducing prediction complexity, proposes a prediction method variable length Markov tree based on context. This method will be the cognition history state information users' perception of the constructed tree state variable order, dynamic changes in forecasting required recently The number of the history of the state, with low complexity and high prediction accuracy, the characteristics of historical short sequences, and can improve the forecasting accuracy of stable environment by updating the state tree cycle. Considering the spectrum detection error to predict the accuracy, to further improve the algorithm, proposed a combination of hidden Markov spectrum model and context tree variable length Markov model prediction method, first channel state Cin Markoff model to restore the real training sequence, then established the state tree, thus eliminating influence of detection error on the performance of the pre measured spectrum. Using the spectral data generated based on queuing model, in a stable environment and non-stationary environment respectively the effectiveness of the context variable length tree based on Markov method were used. Spectrum queuing model and discrete time Markov model using data validation In the case of detection error, the effectiveness of the algorithm combining the hidden Markov model with the variable length Markov model of the context tree is effective.

【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN925

【参考文献】