有結構性之等增益碼書設計及其快速搜尋演算法

Abstract

這篇論文中，我們對於單數據流和多數據流提出有規則性結構的等增益碼 書設計演算法。此外，我們也提出相對應的快速碼書搜尋演算法以降低搜尋最適 當的碼字所需之複雜度。從模擬的結果顯示，所提出的等增益碼書與其他非等增 益碼書之效能比較只有些微差異。模擬結果也顯示出，利用提出的快速搜尋演算 法可達到幾乎沒有錯誤率的良好結果，複雜度卻因碼書本身的結構特性而大幅下 降。

In this thesis, we propose an algorithm to design regularized-structure equal gain codebook for single-stream and multi-stream data. In addition, their corresponding fast codebook search algorithms are also proposed to reduce the complexity for searching the most suitable codeword in a given codebook. From the simulation results we see that our proposed ”equal-gain” codebook has good performances with negligible loss compared with other optimal ”non equal-gain” codebooks but with good structured characteristic, which is convenient for codeword search. The simulations also show that the proposed fast codebook search can achieve a result merely without any loss, and can greatly reduce the computational complexity thanks to their highly regularized code structure.