無失真音訊編碼之整數型參數估測器之研製

近年來，音訊壓縮技術已經日益普及，其中有兩大類的壓縮技術，一個是失真型系統，另一個則是無失真型系統。到目前為止，已經有許多有效率的編碼架構被提出，有些甚至被制定成為標準或規格，然而這些編碼架構多是屬於失真型的系統。但隨著硬體方面的儲存空間大幅增長及網路傳輸頻寬大幅度增加後，無失真型的壓縮系統也慢慢地受到各方的重視。儘管無失真型壓縮系統的壓縮率不能與壓縮型系統來相提並論，但我們所重視的是將原音重現給聽者，相對而言，壓縮對於無失真系統就有點像是一個還不錯的附加價值了。 ALS(Audio LosslesS)編碼的基本流程依順序為先做估測(prediction)再做entropy coding，就估測這部分來說，估測是用於降低鄰近點間的重複部分(redundancy)，而entropy coding是將估測完的估測誤差(prediction error)進行編碼。根據成本考量﹑硬體設計和應用層面，估測器(predictor)可分成兩類：整數型參數及非整數型參數系統，本論文是針對整數型參數系統來設計，而這些參數是事先設計好的(predesign)，所以適用於查表(table look up)方式來決定估測器的參數；另外，我們將使用Rice encoder來當我們的entropy encoder，在這邊，每次整批資料處理大小(block size)的最佳化及Rice code裡的參數L的優化是我們另一個著眼點。實驗結果顯示針對我們的測試音樂(tested materials)而言，依照我們提出的方法所設計的整數型參數組確實是可以符合這些音樂的特性，而且相較於已發表論文中的整數型參數系統，我們的系統在壓縮率方面可以更佳。此外，由於Rice coding是一個不定長度(variable length)的編碼系統，所以我們以不同的音樂去編碼的話，音樂的壓縮率也就會隨之變化。

關鍵字

無失真；音訊；壓縮；整數

並列摘要

In the recent years, the audio compression techniques have become more and more popular, there are two major kinds of techniques, one is lossy scheme and the other is lossless one. So far, many effective compression coding algorithms have been proposed and some of them have been standardized. However, most of them are lossy systems. Accompany with great improvement on storage hardware and network transmission bandwidth, the audio lossless scheme has been paid much attention progressively. In spite of the compression ratio of lossless system is less than lossy system, what we want is the original audio recur completely to listeners, the compression of lossless system is an additional benefit. The basic procedure of ALS (Audio LosslesS) coding system is prediction for first and then entropy coding; about prediction part, the prediction is used for removing the redundancy among adjacent samples, as for entropy coding that is to encode prediction errors. According to costs, hardware constraints and application areas, the design of predictor can be divided into two parts, integer coefficients predictor and non-integer coefficients predictor, in our thesis, the objective is to design the integer coefficients predictor where coefficients are predesigned and can be used in the form of table look up methods; besides, we take Rice encoder for entropy encoder we mentioned above, in this part, the optimization of block size and parameter L are our focus. Our simulation results showed that we indeed proposed an efficient method to design the integer coefficient sets which adapt to our tested materials, and our system can get better performances compare to the same integer coefficients system which has been proposed. And, owing to Rice coding is a variable length coding, the compression ratio will change with different tested materials.

並列關鍵字

audio ； coding ； index table ； integer ； lossless ； predictor

參考文獻

[8] T. Liebchen, “An Introduction to MPEG-4 Audio Lossless Coding,” IEEE ICASSP,

pp. Ⅲ-1012 – 1015, 2004.

[11] R. F. Rice and J. R. Plaunt, “Adaptive variable-length coding for efficient compression of spacecraft television data,” IEEE Trans. Comm., vol. 19, no. 6,

[12] M. Hans and R. W. Schafer, “Lossless Compression of Digital Audio,” IEEE Signal Processing Magazine, pp. 21 – 32, 2001.

[1] X. Lin, W. H. Tang, C. W. Gee, G. Li, “A Lossless Audio Compression Software for Windows Application,” Signal Processing Proceedings, ICSP, Fourth International Conference, Vol.2, 12-16 Oct, 1998.

Google Scholar

國際替代計量

無失真音訊編碼之整數型參數估測器之研製

未授權

主題瀏覽