【作者】 王培雅；

【导师】 张朝阳；

【作者基本信息】 浙江大学 ， 信息与通信工程， 2010， 硕士

【摘要】 认知无线电技术是目前解决频谱资源稀缺问题的核心技术,频谱感知是保证该技术得以实施的前提。为了保证授权用户的正常传输,同时满足认知用户的速率要求,必须要求认知用户能进行多信道频谱感知,以便发现足够的“频谱空洞”进行频谱接入,并监测频谱活动状态的变化。本论文采用置信传播算法作为解决手段,研究了基于随机信道选择和欠采样技术的多信道频谱感知方法,以及异步OFDMA网络的频谱感知方法,对于认知无线电的频谱感知的研究做出了一定的贡献。首先研究了基于随机信道选择的多信道频谱感知问题。阐述了本课题的系统模型,分析了检验统计量的渐近概率分布及本模型的等效二部图形式,并引入二部图的经典融合算法——置信传播算法。随后提出了三种频谱感知算法,其核心思想是认知用户随机选择部分信道进行感知,然后采用置信传播等推理算法判决出所有信道的活动状态。方法一中,仅当融合中心接收到足够多的感知信息后才开始置信推理。方法二中,当融合中心一接收到感知信息就开始推理判决,并结合“信道释放”机制,即不对那些已经判决出活动状态的信道进行重复感知,进一步降低感知延迟。方法三中,认知用户将方法一、二中的感知信息进行量化处理,并选择可靠性较高的感知结果发送给融合中心进行融合判决,降低宽带限制。其次研究了基于欠采样技术的多信道频谱感知问题。阐述了本课题的系统模型及授权用户的时域信号模型,分析了采样速率不满足奈奎斯特采样定律时的频谱混叠方式,并分析了检验统计量的渐近分布及本模型的等效二部图表示。随后提出了一种基于欠采样技术的频谱感知算法,其基本思想是认知用户在每次感知过程中按一定规律选择采样率对整个频谱进行采样感知,并利用置信传播算法对感知信息进行推理判决。最后重点分析了感知算法中的采样率选择问题,证明了认知用户选择授权用户信道带宽的质数倍作为采样速率较好。然后研究了异步OFDMA网络的频谱感知问题,并重点分析了存在分数倍载波频偏时的频谱感知问题。研究了存在分数倍载波频偏时的各子载波上的接收信号模型以及载波频偏对传统感知算法的影响,给出了最差情况下的虚警概率和检测概率的渐近表达式。随后简化了频域信号的接收模型,提出了一种基于置信传播的感知方法,能有效降低或抑制信道间干扰对频谱感知的影响,大大降低虚警的发生。最后对能量检测算法进行改进,提出了简化的感知方法,降低了计算复杂度,易于硬件实现。并基于ADSP-TS201,给出了简化的感知方法的DSP实现方法。更多还原

【Abstract】 Nowadays, cognitive radio is a key technology to solve the problem of spectrum scarceness, and spectrum sensing is the precondition to achieve such technology. In order to guarantee the transmission of licensed users and satisfy the rate requirement of cognitive users simultaneously, multi-channel spectrum sensing is required. So cognitive users can discover enough spectrum holes for spectrum access and can sense the change of spectrum activities. In my paper, I have studied multi-channel spectrum sensing based on random selection of channels, multi-channel sensing based on down sampling, and spectrum sensing in asynchronous OFDMA network, which gives some significant contribution to the research on spectrum sensing.Firstly, the problem of multi-channel spectrum sensing based on random selection of channels is studied. The system model of spectrum sensing is described, the asymptotic probability density fuction of the test statistic and the equivalent bipartite graph are analyzed, and belief propagation algorithm (BPA) is introduced, as a typical fusion algorithm in the bipartite graph. Three spectrum sensing methods are proposed then, whose basic idea is that cognitive users choose part of channels to sense randomly and infer the spectrum activities of all the channels by BPA. In method one, the fusion center starts data fusion only when enough sensing information is received. In method two, once sensing information is received, the fusion center starts data fusion and never detects the channels whose activities are already judged, which can decrease the sensing delay further. In method three, the sensing information in the above methods would be quantified, and only the reliable results would be chosen for data fusion, which can decrease the bandwidth constraint.Next, the problem of multi-channel spectrum sensing based on down sampling is studied. The system model and the signal model of licensed users in time domain are characterized, the signal model of licensed users in frequency domain when the frequency sampling rate doesn’t satisfy the theorem of Nyquist Sampling is investigated, and the asymptotic probability density function of the test statistic and the equivalent bipartite graph of our model are analyzed. Then a kind of spectrum sensing method is proposed, whose basic idea is that the cognitive user chooses the frequency sampling rate according to a certain rule to sense the whole wide-band spectrum in every sensing process and infers the spectrum activities of all the channels in terms of BPA. The problem of the selection of sampling rate is mainly analyzed. And it proves that it is better to choose the frequency sampling rate which is a multiple of the channel bandwidth of licisend users and the value of multiple is a prime.Finally, the problem of spectrum sensing in asynchronous OFDMA network is studied, especially when fractional carrier frequency offset exists. The model of the received signal in each subcarrier and the effect to spectrum sensing induced by CFO are analyzed firstly. Then the signal is simplified, and a spectrum sensing method based BPA is presented, which can decrease the effect to spectrum sensing induced by CFO and false alarm. After that a simplified sensing method is proposed by advancing energy detection, which can reduce the complexity of computation and can be implemented on the hardware device easily. And the DSP implementation of the simplified method is introduced based on ADSP-TS201.更多还原