| 摘要: | 本論文主要研究無失真音訊壓縮法，無失真音訊壓縮可以壓縮音樂訊號，並且能夠完全重建原始訊號而無任何損失。
於無失真音訊壓縮的領域，研究通常可分為兩個發展方向，訊號預測技術與編碼演算法。訊號預測技術用於理解原始訊號，並使用一個模型來描述訊號且去除訊號的相關性，其通常是藉由線性預測模型來進行；編碼則是依據符號的發生機率，對每一個符號碼進行更有效率的二進位碼編製。
在這篇論文裡，我們從無失真壓縮的觀點，針對上述兩個方向進行深入研究。最後並且提出四項設計原則，這四項設計原則指出，如何改進無失真音訊壓縮的壓縮效率。基於這四項設計原則，我們並設計出一個新的無失真音訊壓縮演算法。我們的方法主要包含，分框、去關連、平滑、預測、編碼以及一個適應性最佳壓縮策略。於實際測試結果中顯示，我們的方法具有高壓縮性能。
This thesis studies lossless audio compression algorithms. The lossless audio compression enables the digital audio data encoding without any loss in quality due to a perfect reconstruction of the original signal.
In the domain of lossless compression, research works contain two broad development sections, signal predictive modeling techniques and coding algorithms. The signal predictive modeling techniques are concerned with the understanding of the source signal and utilizing the modeling method to decorrelate a signal. The coding algorithms focus on the processing tasks of efficiently representing a single symbol as a code, usually in binary, given a set of estimated symbol probabilities.
In this thesis, all two sections are investigated in depth and handled from the lossless viewpoint. There are four design principles proposed. The principles aim to improve the lossless audio compression efficiency. Base on these principles, a new lossless audio compression algorithm has been developed to achieve the goal. The algorithm consists of different processing modules including framing, decorrelation, smoothing, prediction, coding, and daptation for optimal compression. Finally, the simulation
results have shown the high compressing performance of the proposed lossless audio compression algorithm. |
