使用多重假設檢定之寬頻感知系統合作式頻譜偵測

Abstract

藉由使用多重假設檢定，對於寬頻感知系統合作式頻譜偵測提出一個較低複雜度的決策融合(decision fusion)方法。為了維持多通道頻譜偵測的偵測品質，利用Benjamini & Hochberg 程序來控制不同環境設定下的系統整體錯誤偵測率(FSR)和錯誤忽略率(FIR)；此外又更進一步提出一個合併FSR 和FIR 的偵測方式，希望能同時達成降低誤失率(missing ratio)和降低誤警率(false alarm ratio)的目標；最後，與其單純偵測許可使用者(licensed user)的存在與否，我們在融合中心(fusion center)根據訊息回報的方式為能量回報(EFE)或是決策回報(DFE)，設計了一個以二維度假設檢定為特色的訊號強度估計程序。 模擬結果顯示出，對於控制錯誤偵測率(FSR)，提出的決策融合方法在誤警率(false alarm ratio)勝過能量融合(energy fusion)方法的同時，能保有媲美能量融合方法的誤失率(missing ratio)。至於控制錯誤忽略率(FIR)，若假設融合中心可以完美獲得各個通道的訊號雜訊比值，則誤失率可以被降低到很低的程度；若是考慮比較實際的情況當融合中心認定系統操作在某一名義上的訊號雜訊比值時，雖然誤失率在訊號雜訊比值低時會被提高些許，但整體表現仍然維持在可接受的錯誤率水準之下。在實施隨訊號雜訊比值調整決策法則之後，合併FSR 和FIR 的偵測方式可同時擁有令人滿意的誤警率和誤失率，為我們的最佳選擇。在訊號強度估計程序的準確性方面，使用能量回報(EFE)比使用決策回報(DFE)來得準確，不過這兩種回報方式對於通道的可利用性皆有著很好的判斷能力。

A lower complexity decision fusion method is proposed for cooperative sensing of wideband cognitive radio using multiple hypotheses testing. To maintain the quality of spectrum sensing in multiple channels, the Benjamini and Hochberg procedure is applied to control the False Sensing Rate (FSR) and the False Ignorance Rate (FIR) of the system under different settings of the environments. Besides, a combined FSR and FIR approach is further proposed to have lower missing and false alarm ratios simultaneously. Instead of detecting the presence of the licensed user, a signal strength estimation procedure which features two-dimension hypotheses testing is also proposed, using the energy feedback (EFE) or decision feedback (DFE) information at the fusion center. In regards of the FSR, simulation results show that the proposed sensing method outperforms the energy fusion method in the false alarm ratio, while maintains a missing ratio comparable to the energy fusion method. As for controlling the FIR, a low missing ratio can be achieved by assuming that ideal SNRs of channels are available at the fusion center. Considering a more practical condition where only a nominal SNR is used at the fusion center, the missing ratio can still show below an acceptable level in spite of losing some performance in the low SNR region. After introducing the SNR switching rule, the combined FSR and FIR approach performs the best and has both satisfactory false alarm and missing ratios. For the signal strength estimation procedure, the EFE has the better accuracy than the DFE, and both of them can achieve a low enough error rate of the availability of the sub-bands.