標題: 室內無線通道匹配之向量量化設計
Robust Vector Quantization for Indoor Wireless Channels
作者: 李中任
Chung-Jen Lee
張文輝
Dr. Wen-Whei Chang
電信工程研究所
關鍵字: 室內無線通道;向量量化;哈達瑪轉換;基因演算法;Indoor Wireless Channels;Vector Quantization;Hadamard Transform;Genetic Algorithm
公開日期: 2000
摘要: 向量量化是目前常見的影音壓縮處理技術,其目的在有效運用有限的網路傳輸頻寬與儲存容量,但通道失真會改變其編碼輸出的碼字索引進而影響通訊品質。理想的系統設計應兼顧其製作即時性與傳輸強健性。一具體可行方案是應用區段碼的對映運算來建構限定向量量化碼書,其關鍵在於訓練階段須配合不同的通道特性慎選其區段碼,同時依據最小量化失真準則設定相關的對映矩陣。目前相關研究主要集中在無記憶性二元對稱通道,並不足以正確反應室內及室外無線通訊環境因多路徑衰退所衍生的叢發位元錯誤。有鑑於此,本論文中先完成室內無線通訊系統之模擬,並量測其傳輸位元序列的通道轉移機率,再配合哈達瑪轉換分群理論的數學推導,進行索引分類據以發展一通道匹配的區段碼選擇機制。有關對映矩陣的配置,主要是重新規劃其組成向量的能量分布成為一非線性參數預估問題,並參考基因法則建立其最佳化訓練的隨機搜尋演算機制。
Vector quantization (VQ) is an efficient approach to data compression of speech and image. However, transmitting VQ data over noisy channels change the index bits and consequently leads to severe distortions in the reconstructed output. This motivates our research into trying to design a zero-redundancy VQ system that achieves high channel robustness by using a constrained codebook through linear mapping of a block code. Design of a constrained VQ codebook involves selecting a good block code as well as an optimization of the mapping matrix for that specific block code. Further robustness can be realized by matching the real channel behavior to the channel model on which the codebook design is based. The strategy applied here for VQ analysis is based on a Hadamard framework, in which block code components that minimize the channel distortion are chosen in accordance with the Hadamard transform of channel transition probabilities. To optimize the mapping matrix to a given block code, its design is formulated as a combinatorial optimization problem that is amenable to the application of real-coded genetic algorithm. For better tracking the intrinsic natures of channel errors, this study integrates the proposed codebook training algorithm into a simulated indoor wireless communication system.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT890435021
http://hdl.handle.net/11536/67302
Appears in Collections:Thesis