Variable tuple size based entropy coding for transform domain audio signal

다국어 초록 (Multilingual Abstract)

As audio compression is one of the core technologies that enable media transmitting and storing within the limited network bandwidth, it is crucial to employ audio codecs to support superior compression performance. As the state-of-the-art audio coding standard, Unified Speech and Audio Coding (USAC) was developed by the ISO/IEC Moving Picture Experts Group (MPEG) audio group. Although with the development of audio coding standards, many coding tools, such as prediction, quantization, and filtering, have been presented, entropy coding tools have only minor changes. In this thesis, to investigate the trade-off between the compression performance and the memory capacity of context tables, we explore the entropy coding methods based on the tuple, a set of transformed quantized audio samples. In addition, we propose Variable Tuple Size Based entropy coding, which is dividing transformed quantized audio samples as multiple variable tuples and a remaining tuple. Experimental results show that the proposed variable tuple coding can achieve improved compression performance as high as 16% while maintaining the memory capacity of context tables compared to the fixed 4-tuple grouping.

목차 (Table of Contents)

• 1. Introduction 10
• 2. Related Works 14
• 2.1 Overview of Entropy Coding 14
• 2.2 Entropy Coding for Audio Signal 26
• 2.3 Entropy Coding for Video Signal 37
• 3. Proposed Methods 45
• 3.1 Tuple Based Entropy Coding 45
• 3.1.1. Fixed N-tuple Grouping 47
• 3.1.2. Adaptive N-tuple Grouping 49
• 3.1.3. Tuple Reuse Flag for N-tuple Grouping 51
• 3.1.4. Multiple Context Tables for N-tuple Grouping 56
• 3.2 Variable Tuple Size Based Entropy Coding 59
• 4. Experimental Results 67
• 4.1 Experimental Environments 67
• 4.2 Experimental Results 69
• 4.2.1. Tuple Based Entropy Coding 69
• 4.2.2. Variable Tuple Size Based Entropy Coding 74
• 5. Conclusions 80
• References 82
• 국문초록 85
• Curriculum Vitae (in Korean) 86