音訊無失真壓縮之整係數估測器設計與實現

DESIGN AND IMPLEMENTATION OF INTEGER COEFFICIENT PREDICTOR FOR AUDIO LOSSLESS CODER

曾世忠(Shih-Chung Tseng)

指導教授 ： 李清坤

大同大學/電機資訊學院/通訊工程研究所/碩士(2006年)

摘要

中文摘要
近年來，由於科技的快速發展，使得人們對各類音訊的品質要求越來越高，在數位化的時代裏，享受原音重現的優質音樂(如CD、DVD MPEG-4)，成為基本的需求。然而，數位化高品質的音樂，意謂著需用更多的位元來重現原音，資料量增加相對的傳輸時間勢必增長，如何運用有限的資源，在不影響音質的前提下，提昇傳輸的速率呢？首先想到的方法就是：減少音訊資料量的大小，也就是進行所謂音訊無失真(ALS) 的壓縮，運用降低資料量來加速傳遞的效率。
音訊壓縮的技術種類繁多，一般常用的壓縮方法，是音訊透過估測器，來模擬原始音訊，並設法降低估測所產生的誤差，再利用Entropy Coding對誤差值進行適當的減化編碼，來達成音訊無失真(ALS) 的壓縮效果。
本篇論文使用ADPCM Pre-designed預先設計估測係數組合的方式，將足量的各類型音訊樣本，透過我們的設計：運用音訊波形各相鄰點之間，斜率與階數的相關性研究，找出參數的合理範圍，並對範圍內所有的係數組合做全面性的比較，選出最佳的係數組合，作為估測器的索引係數 (Index Coefficients)，同時為降低複雜度與加速估測器的運算，我們選用整數型的係數組合；另外在Entropy Coding編碼方面，我們採用一般常用的Rice Coding來進行編碼。
最後我們完成音訊無失真壓縮的程式，並且與網路上所蒐尋到同樣使用整係數估測器的設計及一般常用的壓縮軟體進行比較，結果都有較佳的表現，顯示我們所設計的方法，是一種有效率的音訊無失真壓縮設計。

關鍵字

音訊無失真壓縮 ； 整係數估測器 ； 估測器 ； 壓縮

並列摘要

ENGLISH ABSTRACT
In recent years, the fast development of science and technology. Making people require higher and higher to the quality of all kinds of audio. Enjoy the high-quality music that the original sound reappears in an era of digitization (such as CD, DVD MPEG-4 Audio), become the basic demand. However, the high-quality music of digitization, purpose needs to reproduce the original sound with more bits. The materials increase relative transmission time and certainly will increase. How to use limited resources, on the premise of not influencing quality, promote the speed transmitted? The method to expect at first is: Reduce the size of the materials quantity of the data, that is to say that there is audio lossless (ALS ) compression, use the efficiency transmitting of the amount of materials decreased with higher speed.
The technology that the audio compresses is various in style, the generally commonly used compression method, it is the audio that imitates the primitive audio through estimating the predictor, and try to reduce the prediction error, then utilize Entropy Coding to carry on proper reducing and melting the code with error value again. At the last have to reach the result of audio lossless (ALS) compression.
This paper thesis uses ADPCM with pre-designed predictor coefficients, all kinds of sufficient audio sample in advance, pass through our design: Use it between every adjoin point of wave form of the audio, the dependence that the slope and order are studied, find out the rational range of the parameter, and make to all coefficient in the range up and do comprehensive comparison, then select the best coefficient association, as the index coefficient of estimating the predictor. At the same time in order to reduce complexity and operation estimating the predictor with higher speed, we select the integer coefficient association; In addition, in Entropy Coding code, we adopt generally commonly used Rice Coding to encode.
Finally, we finish the audio lossless compressed program, and find whole design that estimates the predictor of integer coefficient from internet and generally commonly used compressed software to compare with our system. The results all have better behavior. Show the method that we designed. It is a kind of efficient method of the audio lossless compression.

並列關鍵字

ALS ； Audio lossless coder ； Predictor

參考文獻

[3] K. Konstantinides, “An Introduction to Super Audio CD and DVD-Audio,” IEEE Signal Processing Magazine, vol. 20, no. 4, pp.71-82, 2003.

[4] M. Hans and R. W. Schafer, “Lossless Compression of Digital Audio,” IEEE Signal Processing Magazine, vol. 18, no. 4, pp. 21 - 32, 2001.

[10] DVD specifications for read-only Disc, Version 1.0, DVD Forum, 1999 [Online]. Available: http://www.dvdforum.com.

[12] W. C. Chu, Speech Coding Algorithms, New Jersey, May 2003.

[14] A. Gersho and R. M. Gray, Vector Quantization and Signal Compression, Norwell, Kluwer Academic Publishers, Boston, 1992.