Efficient audio coding with optimized subband configurations

Abstract

A dynamic subbands design selection scheme for perceptual audio coding is presented which optimizes the coding efficiency in association with an improved utilization of the masking potential of the signal. The choices for subbands design selection include critical bands, uniform subbands, uniform subbands with a non-uniform last band and a non-uniform-non-critical (NUNC) subbands distribution. A particular consideration is given to corresponding critical band boundaries while formulating NUNC subbands, to minimize the perception of coding noise. The design selection decision is made by first identifying the best tradeoff between empty quantizer slots and amount of side information, followed by a minimization of the perceptual entropy (PE) estimates. A further analysis is carried out to identify the prominent spectral peaks (PSPs), with a view to efficiently utilize their masking potential by the bit allocation algorithm which makes a bit cut first to subbands with PSPs. Test results based on ITU-R recommendation BS.1116 show that our coding scheme performs slightly better as compared to MPEG-4 AAC verification model (VM) for a majority of signal types from the SQAM database. The paper concludes with a discussion of future research implications of the work