標題: MP3聲訊編碼器之研究及簡化設計
Investigation and Simplified Design of MP3 Audio Encoder
作者: 楊智凱
Chih-Kai Yang
陳紹基
Sau-Gee Chen
電子研究所
關鍵字: MP3;MPEG;Audio;Encoder;MDCT;Subband;Bit allocation;Noise shaping;psychoacoustic;Huffman;layer III
公開日期: 2002
摘要: 本篇論文提出兩種改良式的搜尋方法,可以應用在MP3聲訊編碼器中的位元分配迴圈中。MP3編碼器的位元分配演算法,是以迴圈的方式來尋找最適合的量化階數,每執行一次迴圈,都必須重新對經由改良式餘弦轉換,所得到的頻譜係數做量化和編碼的運算,因此迴圈執行次數的多寡,就決定了編碼器的性能。為了簡化MP聲訊編碼器的複雜度,我們提出兩種改良式的搜尋方法,分別適合應用在批次處理和即時處理的系統上。和Lame編碼器相比,我們的第一種改良式搜尋法,可以減少1%~30%的迴圈執行次數;和Dist10編碼器相比,我們的第二種改良式搜尋法,可以減少15%~70%的迴圈執行次數。
In this thesis, we propose two improved search algorithms for bit allocation loop in MP3 audio encoder. The goal of bit allocation loop in MP3 audio encoder is to find the smallest step size for non-uniform quantizer. Each iteration loop will perform one quantizer and Huffman coder routine. And the performance of MP3 audio encoder is depending on the counts of iteration loop. One algorithm is for batch processing system and the other is for real-time application. Compared to the first-order search algorithm in Lame MP3 encoder, the first improved search algorithm can reduce loop counts from 1% to 30%. Compared to binary search algorithm in Dist10 MP3 encoder, the second improved search algorithm can reduce loop counts from 15% to 70%.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT910428008
http://hdl.handle.net/11536/70350
顯示於類別:畢業論文