合作式通訊下之分散式渦輪碼

Abstract

文獻指出分散式渦輪碼(Distributed Turbo coding)可以接近無線中繼網路(Wireless Relay Network)的通道容量上限(Capacity)。實際的分散式渦輪碼，如果在中繼端發生故障(Outage)將轉換至非合作的模式，因為如果繼續中繼錯誤的資訊將會誤導目的地終端，並造成錯誤的傳遞繁殖；而此非合作模式將使系統效能下降。最近，一種轉送模式稱之為解碼放大轉送模式(Decode-Amplify-Forward)被提出來解決這個問題。不像分散式渦輪碼，解碼放大轉送模式既不在中繼端重新編碼，也不在目的地終端做遞迴解碼，因此擁有較有效率之運算複雜度。然而，解碼放大轉送模式的中繼端需要知道Log-Likelihood Ratio的分佈，一般被模擬成高斯混合(Gaussian Mixture)。在本論文裡，我們提出使用EM algorithm做為識別高斯混合的方法，模擬結果顯示EM algorithm可以更準確地識別此分佈，因而解碼放大轉送模式的效能也得到提升。而分散式渦輪碼可與解碼放大轉送模式合併為一種混合轉送模式，此一模式比其他我們所比較的分散式渦輪碼的變形都擁有較好的效能。

It has been shown that distributed turbo coding (DTC) can approach the capacity of a wireless relay network. Practical DTC has to switch to non-cooperative transit mode should inter-user outage occur, since erroneous bits, if (re-encode and) forwarded to the destination, will mislead the destination and cause severe error propagation. This non-cooperative transit mode degrades the performance. Recently, a forwarding strategy termed decode-amplify-forward (DAF) is developed to solve the problem. Unlike DTC, DAF neither conducts encoding at the relay nor iterative decoding at the destination. It is superior and computationally more efficient. However, DAF requires the relay to have the likelihood-ratio (LLR) distribution, typically modeled as a Gaussian mixture. In this thesis, we propose to use the EM algorithm for the identification of the Gaussian mixture. Simulations show that the EM algorithm can identify the LLR distribution accurately and the performance of DAF can be enhanced. Further, DAF can be combined with DTC resulting in a DTC-DAF scheme. We have also studies various DTC related schemes and found that DTC-DAF can have the best performance.