Full metadata record
DC FieldValueLanguage
dc.contributor.author梁家銘en_US
dc.contributor.authorLiang, Chia-Mingen_US
dc.contributor.author劉啟民en_US
dc.contributor.authorLiu Chi-Minen_US
dc.date.accessioned2014-12-12T02:15:04Z-
dc.date.available2014-12-12T02:15:04Z-
dc.date.issued1995en_US
dc.identifier.urihttp://140.113.39.130/cdrfb3/record/nctu/#NT840392012en_US
dc.identifier.urihttp://hdl.handle.net/11536/60353-
dc.description.abstract格狀向量量化法是一種均勻量化法, 所以可適當控制量化誤差的範 圍以配合 由聲響心理模型所求出來的遮蔽門檻。 本篇論文主要是探討將 格狀向量量化法應 用於音訊編碼, 同時評估格狀向量量化法在極低壓縮 位元率音訊編碼這方面的潛 力。 由於 MPEG 是一個國際的音訊壓縮標準 , 以它當做是主架構比較的參考標 準會較有說服力, 所以我們選擇 MPEG layer I的架構做為所提出來的編碼器的主 要骨幹。 我們將會說 明如何在一已知維度的空間中設計一個最佳格狀向量量化 器的具體步驟 , 同時我們也會提出這個最佳格狀向量量化器經由式子所推導出來 的最 佳理論值有多少, 以做為評估實驗結果的一個參考。 除了驗證格狀向量 量 化器在幾何上具有較緊密的優點外, 我們將對格狀向量量化器做些 修正, 使其 能夠利用到經正規化後之輸入向量本身的不均勻分佈之特 性, 進一步得到更好的 效能。 從實驗結果分析顯示, 在 32K 的位元 壓縮率下, 格狀向量量化器幾何上 較緊密的優點提供約平均 1.36 dB 的改善, 而且考慮經正規化後之輸入向量本身 的不均勻分佈之特性後 可以另外獲得 0.18 dB的改善量。 因此, 格狀向量量化器 在極低壓縮 位元率音訊編碼這方面有很大的應用空間。 Lattice vector quantizer is a uniform quantizer, and it can conform the quantization error to the masking threshold computed from the psychoacoustic model analysis. In the thesis, we attempt to apply lattice vector quantization (VQ) to audio coding and investigate the potential of lattice VQ in very low bitrate coding. The encoding structure of MPEG layer I is chosen as the backbone of our proposed coder because MPEG is a inter-national audio coding standard and it will be persuasive when it is taken as a benchmark. We will demonstrate the procedure to design an optimal lattice VQ based on the root lattice in a given dimension. The theoretical benefit of lattice VQover scalar quantization will be analyzed and verified through experiments. In addition to verifying the geometric compactness of lattice VQ, we also make proper adjustment to the lattice VQ for taking advantage of the nonuniform distribution of normalized input vectors. The experiment results show that at 32 kbps, we can obtain average 1.36 dB gain from the geometric compactness of lattice VQ and additional 0.18 dB from the nonuniform distribution of normalized input vectors. Therefore, it is convincible that lattice VQ is applicable in very low bit rate audio coding.zh_TW
dc.language.isozh_TWen_US
dc.subject格狀向量量化zh_TW
dc.subject音訊編碼zh_TW
dc.subjectAudio Codingen_US
dc.subjectLattice Vector Quantizationen_US
dc.subjectMPEG Layer Ien_US
dc.title在 MPEG Layer I 的架構下將格狀向量量化法運用於音訊編碼zh_TW
dc.titleApplying Lattice Vector Quantization to Audio Coding in the Framework of MPEG Layer Ien_US
dc.typeThesisen_US
dc.contributor.department資訊科學與工程研究所zh_TW
Appears in Collections:Thesis