可變長度碼的疊代訊源通道解碼機制

Abstract

本論文主要探討基於可變長度碼之索引層級疊代訊源通道解碼機制，及其在心電訊號壓縮的應用。傳統的通道解碼演算法受限於其位元層級處理模式，與基於索引層級而推導的訊源解碼演算法存在著相容性問題。有鑑於此，我們首先利用可變長碼在索引層級上的分段籬柵結構，結合通道編碼器之狀態轉移以建構由二維狀態組成的籬柵圖。接著依據此籬柵圖推導其軟性通道解碼演算法，並將其整合運用於疊代解碼機制。主要是針對具特定機率分佈的訊號源，利用其可變長編碼所得的殘餘冗息，再基於渦輪碼原則在疊代解碼過程中相互交換訊源及通道解碼器輸出的額外訊息。實驗模擬証實，索引層級的疊代解碼機制能有效提升系統的強健性。

Transmission of convolutionally encoded source codec parameters over noisy channels can benefit from the use of iterative source‐channel decoding (ISCD). This thesis focuses on the design of a symbol-based ISCD for variable‐length code (VLC). Our work starts with a sectionalized code trellis whose two-dimensional states are created by combing the symbol-based VLC trellis and the state transition of a channel encoder. We first derive a soft-output channel decoding algorithm that integrates the residual source into the ISCD algorithm. Based the turbo principle, ISCD exchanges the extrinsic information resulting from the source decoder and the channel decoder to improve the decoding performance. Simulation results are presented for Gauss-Markov sources and ECG signals which demonstrate the error-resilience capabilities of symbol-based iterative decoding.