Trellis-based optimization of MPEG-4 advanced audio coding

Abstract

We outline a method to perform efficient low rate quantization for MPEG-4 advanced audio coding (AAC). The AAC bit stream consists of indices for quantized spectral coefficients as well as side information about quantizer step sizes and Huffman codebooks. The MPEG-4 Verification Model does not explicitly account for side information bits in its optimization and suffers from poor compression efficiency at low bit rates. We reformulate the encoding problem as one of optimal parameter selection, where the side information bits are taken into account, so as to minimize the noise to mask ratio for the given target bit rate. The optimal solution is determined by a dynamic programming procedure that efficiently searches through a trellis. This trellis-based optimization greatly improves the low bit rate performance of AAC and, consequently, the performance of a multi-layer AAC system. The resulting bit stream is standard-compatible, and additional complexity due to the proposed optimization is only incurred at the encoder.