An Efficient Implementation of MPEG2 AAC Decoder on SCREAM DSP-16

• Main Authors: Hao Chin, 秦昊
• Other Authors: Wen-Yu Su
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/08817310588502274157

碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 94 === Recently, audio coding schemes based on principles of psycho-acoustic provide superior coding efficiency. They help the storage and transmission of audio information a lot, and also make related electronics popular. MPEG-2 AAC is a standard defined by MPEG(Moving Pictures Expert Group) in 1997, it provides an outstanding coding efficiency. There is wide demand for audio decoder built around embedded systems besides applications on personal computers. In the most case, this kind of consumer electronics use a DSP(Digital Signal Processor) to speed up the numerical computation essential to decoders. SCREAM DSP-16 is a DSP developed by SCREAM LAB of National Cheng-Kung University, it supports some specialized instructions suitable for audio-related numerical computation. In this thesis, several efficient techniques are proposed to implement MPEG-2 AAC decoder on DSP-16. To design an efficient solution on DSP, computation complexity and memory utilization are major concerns. To reduce computation complexity, we use fixed-point operations to replace floating-point operations, then in some critical functions we rewrite the source code in assembly language to further reduce complexity. In addition, this thesis describes several approaches to improve memory utilization.