本論文主要是探討語音壓縮規格中,低位元率的ITU-T G..729語音編碼器之化簡研究為主。 我們在探索ITU-T G..729語音編碼器的過程中發現,在G.729的語音編碼運算處理過程中,固定碼簿的搜尋過程就佔了運算處理約20%的計算量,也因此,我們希望能在維持一定的語音品質之前提下,找出最佳的搜尋方法,將這部分的計算量加以減少。 實驗數據得知,傳統固定碼簿需搜尋8192次,而本論文提出之方法最低僅需2次,節省計算量約98.54%,而聲音品質改變並不多,MOS値從4.5降至3.8,因此,實驗數據證實我們提出之固定碼簿搜尋簡化法效能不錯。
The computation improvement on G.729’s fixed codebook is researched and implemented. Moreover then 20% computation of G.729 is paid on the fixed codebook search. In this thesis, an preprocessing algorithm is employed to limit the search domain of fixed codebook’s space. In the experimental results, more than 98.54% of computation is reduced. The speech quality is still good in this approach. The MOS of original G.729 is 4.5 and the new approach is about 3.8. It is reasonable good and the performance of this approach is confirmed.