| 摘要: | 隨著無線通訊服務快速蓬勃發展，資料傳輸量與頻寬需求量持續倍增，現有靜態頻譜分配方式已無法負荷未來通訊應用服務發展。感知無線電(Cognitive radio, CR)為一新興技術，藉由即時偵測未被充分利用之空頻帶，提供給次用戶(Second user, SU)更多動態存取頻譜(Dynamic Spectrum Access, DSA)的機會，有效改善頻寬利用效率。然而在滿足奈奎斯取樣定理(Nyquist sampling theorem)之下，感知無線電對於偵測稀疏頻帶(Sparse spectrum)之效率不高外，偵測複雜度將隨著所偵測的頻寬變大而增加；再者無線通道產生的遮蔽效應(Shadow fading)也將大幅降低頻譜偵測準確性。
於本論文中，我們設計一寬頻感知正交分頻多工系統(Orthogonal frequency division multiplexing, OFDM)，藉由壓縮感測(Compressive sensing, CS)技術來減少偵測所需取樣點與改善頻譜偵測效能。在貝氏架構(Bayesian framework)下，將問題透過因子圖(Facto-Graph, FG)來解析，並經由可信傳遞(Belief propagation, BP)演算法以疊代方式趨近於全域最佳解。藉由信號之稀疏特性，我們提出子區塊分割法(Sub-block partition method )來簡化信息傳遞法(Message passing algorithm)在數學上之積分問題。
此外，由於訊號具有空間分集特性(Spatial diversity)，因此次用戶基地台可利用分散式合作通訊(Distributed cooperative communication)將偵測主用戶(Primary user, PB)頻帶使用結果，同樣經由信息傳遞與其他次用戶基地台(Secondary base station)間彼此交換訊息，以多次遞迴方式來判定空洞頻帶，有效改善感知無線電於嚴重遮蔽效應下之頻譜偵測效能。
With the growth in wireless services, the demand for high data rate wireless multimedia applications has been significantly increasing recently. The current static spectrum allocation policy is inefficient, and the spectrum utilization of some outdated licensed bands is even more spare.
Cognitive Radio (CR) is emerging technology to improve the efficiency of spectrum utilization by allowing other unlicensed users to utilize unoccupied spectrum holes through Dynamic Spectrum Access(DSA). However, according to the Nyquist sampling theorem, to sensing the sparse spectrum is inefficient, and the computational complexity and cost will also rise in wideband spectrum sensing. Moreover, in urban areas, it is a major challenge to reliably sense the vacant spectrum due to the hidden terminal problem and severe shadow fading.
In this thesis, we design an orthogonal frequency division multiplexing(OFDM) system for compressive spectrum sensing that can sample sparse signal under sub-Nyquist rate to ease the complexity. Under the Bayesian inference framework, we apply an efficient algorithm belief propagation(BP) that is proposed to solve the Bayesian maximum-a-posterior(MAP) inference problem by sequentially and iteratively passing messages. Since the signal has sparse property, we proposed a sub-block partition method to simplify the mathematically integral problem in message passing algorithm。
Furthermore, in our system model, multiple secondary base-stations(BSs) can mitigate the uncertainty of single detection by distributed cooperative spectrum sensing due to spatial diversity. All base stations can communicate through the backhaul by passing message iteratively to overcome the severe shadowing and improving the spectrum sensing accuracy. |
