M/S 編碼技術於 MP3 與 AAC 上之改良

Abstract

M/S編碼技術主要應用在去除訊號的多餘度，進而有效率的使用位元數來達到最大的編碼效益。在音訊壓縮理論裡，如果我們想要無失真的去編碼左右聲道，其代價往往高於同時編碼兩個單聲道的音訊。因此，如何利用雙聲道間的關連性來降低位元率的負擔便是一個主要的課題。而本篇論文所探討的M/S編碼技術便是應用在關連性移除的一種工具，它就提供我們一種矩陣的轉換，讓我們可以從不同的座標軸來看待訊號，從中減少聲道間的相依程度。此外，M/S編碼技術應用在頻帶或頻道的層次也會影響我們使用時機上的決定。

所以要如何成功的設計M/S編碼技術？我們將其歸納為四點，分別為M/S編碼模式切換的時機，M/S聲道的心理聲學模式，聲道間的位元分配與最後聲道內的位元分配等議題。在論文中，我們考慮了以上四點爭議並提出有效率的M/S編碼方法於實際的編碼器。對於頻帶層次的M/S編碼部分，我們更提出Viterbi演算法來改進編碼器的速度與音訊輸出的品質。在品質的量測上，除了主觀的聆聽評量之外，我們還採用了ITU所發展的PEAQ測試系統來評估音訊壓縮後的誤差程度。

M/S (Middle/Side) coding provides an effective method to extract the stereo redundancy. The M/S processing is employing in channel pair signals and using the matrix signal operation to reduce the channel irrelevances, retrieve more available bits back and control the noise imagine well. Furthermore, M/S coding can be applied to more flexible manner in frame based or scale factor band bases. However, the success of this module relies on the careful design in four aspects: M/S usage decision, psychoacoustic model for M/S channels, bits allocated between the M/S channels and bit allocation in the M/S channels. This thesis presents an efficient M/S method designed from those aspects in frame based M/S coding. Furthermore, the Viterbi algorithm is proposed to optimize the scale factor band M/S coding. Both subjective and objective tests have been conducted to demonstrate the improved quality and efficiency over the existing M/S coding methods. The objective system adopted is the PEAQ which is the recommendation system by ITU-R Task Group 10/4.