动态频谱认知无线通信关键技术研究

发布时间：2018-11-18 17:20

【摘要】：认知无线电的关键技术可以概括为频谱感知、频谱共享和频谱管理三个方面。本文重点研究频谱共享方面的星地协同频率选择技术,频谱管理方面的频谱数据压缩、频谱数据挖掘和混沌序列预测技术,最后设计了一种卫星移动通信场景中的动态频谱协同认知无线通信系统架构。本文工作主要有以下四个方面： 1、针对认知无线通信中感知频谱信息交互存在瓶颈的问题,本文提出了一种适用于感知频谱的数据压缩技术,大幅缩减了回传的数据量。通过分析频谱数据的特性和应用传统数据压缩技术存在的缺陷,本文在DCT变换的基础上,通过能量检测结果将频谱数据划分为噪声和信号两部分,并分别采用不同的压缩方案,提高了压缩效能。在此基础上,本文继续深入分析频谱特性,根据频带的邻域相似性提出一种基于信号识别的分段频谱数据压缩算法,改善了DCT变换的低频能量聚焦性,提高了压缩比。仿真实验结果表明分段压缩在绝大部分场景下能同时带来压缩比和失真率两方面的增益。 2、本文在分析频谱数据特性的基础上,针对海量频谱数据挖掘复杂度大的问题提出了一种基于增量运算的频谱数据挖掘方法。运用增量运算的思想,将新回传数据的挖掘信息与已有信息库进行融合,不需要每次都进行全局挖掘,有效降低了运算量。针对频谱数据具有多维特性、稀疏性、非连续性和多变性的特点,本文所提挖掘方法对多项信道质量指标进行了统计分析,提取频谱变化的规律性信息。通过利用预测技术,从挖掘所得信息中生成可靠的频率图谱辅助卫星进行频率选择决策。 3、深入研究了混沌时间序列预测技术,在传统支持向量机预测技术的基础上,设计了3种场景下利用数据特性优化预测模型的技术方案。首先研究了理论混沌系统时间序列预测技术,提出了一种基于迭代误差补偿的LSSVM混沌时间序列预测算法,算法的预测精度相对现有算法提高一个数量级以上。其次对随机性较强的小尺度网络流量预测技术进行了研究,提出了一种基于相关分析的局域LSSVM小尺度网络流量预测算法。算法通过相关分析优化预测模型训练集,有效提高了预测模型的预测精度,并减少了运算量。最后研究了规律性较强的电力负荷预测,提出了一种基于K-means分类的电力负荷LSSVM预测算法,取得了较好的多步预测效果。 4、在卫星移动通信的场景下,提出一种星地协同的认知无线通信系统架构设计方案。卫星认知终端与控制中心协同认知,系统利用控制中心丰富的软硬件资源对终端回传的海量频谱数据进行数据挖掘得到频谱变化的规律性信息,结合认知终端即时感知的频谱环境数据,实现了认知功能的智能化。
[Abstract]:The key technologies of cognitive radio can be summarized as spectrum sensing, spectrum sharing and spectrum management. This paper focuses on the space-ground cooperative frequency selection technology in spectrum sharing, spectrum data compression, spectrum data mining and chaotic sequence prediction in spectrum management. Finally, a dynamic spectrum cooperative cognitive wireless communication system architecture in mobile satellite communication scene is designed. The main work of this paper is as follows: 1. Aiming at the bottleneck of spectrum information interaction in cognitive wireless communication, this paper proposes a data compression technology suitable for sensing spectrum, which greatly reduces the amount of data returned. By analyzing the characteristics of spectrum data and the shortcomings of traditional data compression technology, this paper divides the spectrum data into two parts, noise and signal, on the basis of DCT transform, and adopts different compression schemes. The compression efficiency is improved. On this basis, this paper further analyzes the spectrum characteristics, and proposes a segmented spectrum data compression algorithm based on signal recognition according to the neighborhood similarity of the frequency band, which improves the low frequency energy focusing of the DCT transform and increases the compression ratio. Simulation results show that segmented compression can bring both compression ratio and distortion gain in most scenarios. 2. On the basis of analyzing the characteristics of spectrum data, a method of spectrum data mining based on incremental operation is proposed in this paper. By using the idea of incremental operation, the mining information of the newly transmitted data is fused with the existing information base, and the global mining is not needed every time, which effectively reduces the computation cost. Aiming at the characteristics of multi-dimension, sparsity, discontinuity and variability of spectrum data, the mining methods proposed in this paper are used to analyze the channel quality indexes and extract the regular information of spectrum variation. By using the prediction technique, a reliable frequency map aided satellite is generated from the information obtained from the mining to make the frequency selection decision. 3. The chaotic time series prediction technology is deeply studied. Based on the traditional support vector machine (SVM) prediction technology, the technical scheme of optimizing the prediction model using the data characteristics under three scenarios is designed. Firstly, the time series prediction technology of theoretical chaotic system is studied, and a LSSVM chaotic time series prediction algorithm based on iterative error compensation is proposed. The prediction accuracy of the algorithm is more than one order of magnitude higher than that of the existing algorithms. Secondly, the small scale network traffic prediction technology with strong randomness is studied, and a local LSSVM small scale network traffic prediction algorithm based on correlation analysis is proposed. The algorithm optimizes the training set of prediction model by correlation analysis, which can effectively improve the prediction accuracy and reduce the computation cost. In the end, a new power load LSSVM forecasting algorithm based on K-means classification is proposed, which has a good multi-step forecasting effect. 4. In the scenario of satellite mobile communication, a design scheme of satellite-ground cooperative cognitive wireless communication system is proposed. The satellite cognitive terminal and the control center cooperate in cognition. The system uses the abundant software and hardware resources of the control center to mine the massive spectrum data returned by the terminal to obtain the regular information of the spectrum change. The intelligence of cognitive function is realized by combining the spectrum environment data of cognitive terminal instant perception.
【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN925

【参考文献】