HIGH QUALITY SCALABLE STEREO AUDIO CODING

This paper proposes an efficient, low complexity, scalable audio coder based on a combination of two embedded coding algorithms: the SPIHT (set partitioning in hierarchical trees) coding algorithm [1] and an embedded, nested binary set partitioning (NBSP) algorithm. The SPIHT algorithm, considered to be the premier state-of-the-art algorithm in still image compression, is used for the low frequency subbands in a wavelet packet audio signal decomposition, while the NBSP algorithmencodes the high frequency audio subbands. Both left and right channels are encoded together to form a single embedded stereo audio bitstream, that can be truncated at any point to produce an optimal lower rate and quality bitstream for delivery to lower quality user services. Using standard MPEG test materials, we evaluate the performance of the proposed encoder compared to the MPEG II standard audio coder through informal listening tests at bit rates of 48Kbs/sec and 64Kbs/sec per channel. We conclude that our coder is comparable with MPEG II at 48Kbs/sec and better at 64 Kbs/sec per channel. The algorithm also features exact bit rate control, progressive transmission and low complexity for both the encoder and decoder. These features show its potential for interactive audio transmission over networks.