行動通訊錯誤緩和機制之研究

Abstract

合併音源-通道解碼器的研究，主要是在改良向量量化系統的解碼器，加強其錯誤隱蔽功能以維持位元傳輸的強健性。向量量化編碼器經由樣本碼書比對產生一組最近似輸入信號的碼字，再傳送其索引值作為編碼輸出。而解碼器的訊號還原則是依據接收到的索引值查表取得對應碼字。問題是無線通道環境中存在諸多雜訊的干擾，改變碼字索引的接收值進而造成訊號還原的失真。本論文先針對行動通訓錯誤進行模型分析，以通道匹配技術估測馬可夫模型參數，並結合輸出碼字索引序列的殘餘冗息以提升解碼效能，其關鍵在於配合不同的通道錯誤屬性估算所有可能傳輸索引值的後驗機率，同時依據最小均方誤差準則還原其輸出訊號。我們使用行動通訊平台模擬GSM通訊環境，進行有限狀態馬可夫通道的模型分析，再分別發展基於索引層級和基於位元層級解碼演算法。自迴歸訊號源的模擬結果顯示，基於索引層級演算法較複雜但有較好的系統效能，基於位元層級演算法則易於實現但效能較低。最後，我們整合兩者優點而開發一混合層級的解碼演算法。

This study is concerned with the joint source-channel decoding algorithms having higher robustness against bursty channel errors. The proposed decoder exploits the residual source redundancy and the channel memory information through the use of minimum mean-square error (MMSE) estimation. It is important to address the importance of matching the real channel behavior to the channel model on which the a posteriori probability computation is based. This task was accomplished by using the exponential curve-fitting and the iterative gradient-descent techniques to characterize the channel with memory. We use a mobile communication platform to simulate GSM system and develop the index-level and the bit-level MMSE decoding algorithms. Simulation results indicate that the index-level decoding algorithm performs better than the bit-level decoding algorithm, but at the cost of increased complexity.