用於無線通訊渦輪解碼之適應性運算量控制

Abstract

渦輪解碼是一種利用遞迴的方式來完成解碼的動作，而解碼次數與通道的傳輸品質有相當大的關係，當通道品質相當好時，渦輪解碼可以很快的解碼出正確的資料，然而通道環境惡劣時，渦輪解碼幾乎都得解到所預設的最大解碼次數才會停止。而與其最後才發現解碼是失敗的，到不如及早停止解碼。因此我們提出一個能夠估測封包解碼失敗的方法，並且在盡可能不影響錯誤率的前提下，盡快停止無意義的遞迴動作，以節省不必要的功率消耗，並能縮短解碼時間，提早重傳。其中我們利用一個簡單的方法觀察解碼資料事後機率，來做為判斷是否要放棄解碼的機制，我們稱此機制為預棄技術(Early Give-up)。

另外，基於重傳的封包資料相同的情況下，我們可以利用上一筆解碼失敗所留下來的資訊，做為新的解碼開始的初始值，使得整體平均的解碼遞迴次數能夠再下降，根據模擬結果，結合以上兩種方法，在吵雜的通道環境下我們可以省下最多約80%的遞迴次數。

Turbo Decoding is a kind of iterative decoding process.The number of iterations required to correctly decode the packet depend on the channel condition.Turbo Decoding can obtain correct information quickly when the channel condition is good enough.When the channel is noisy, turbo decoding always needs a lot of the number of iterations which is close to the preset maximum iterations.So we propose a method which can predict that the packet will not be decoded correctly,giving up the decoding process earlier and request data to be re-transmitted immediately. We judge whether the packet is a failed pattern based on observations from the average absolute value of a posteriori probability of turbo decoding. The techbique called “Early Give-up”.

Besides, we can reuse the prior MAP information of the failed process based on the assumption of correlation between same packets transmitted at different times. Simulation results shows that the average iterations required to decode a packet can be reduced up to 80% under bad channel conditions.