Novel Non-linear Inverse Quantization Algorithm and its Architecture for Digital Audio Codecs

Abstract

This paper proposes a novel inverse-quantization (IQ) algorithm and specialized DSP instructions for IQ algorithms. Non-linear inverse quantization is essential in both MPEG-1 Layer-3 and MPEG-2/4 Advanced Audio Coding (AAC). The proposed algorithm is optimized for the non-linear inverse quantization. The proposed algorithm has moderate computational complexity and can achieve the minimum average error. Performance comparisons show a significant improvement compared with existing algorithms. Furthermore, we propose the specialized DSP instructions and their architecture that can efficiently support the proposed IQ algorithm and other IQ algorithms. The audio-specific instruction-set processor that we have previously implemented employs these instructions and their hardware.

Low-Power System Design for MPEG-2/4 AAC Audio Decoder Using Pure ASIC Approach

“…There is a big range of with 8192 entries; therefore, the implementation using table lookup is rather impractical due to memory constraint. A compromised approach is adopted which combines table lookup and interpolation methods [19]- [21]. In order to determine the optimal size of the lookup table with the consideration of cost and quality, we have analyzed the input occurrence property of AAC songs.…”