Enhanced wavelet based audio coder

Several enhancements on the wavelet based audio coder proposed by Sinha and Tewfik (see. Proceedings of the Int. Conf. of Acoust., Speech and Signal Proc. Society (ICASSP), vol.1, p.197-200, 1993) are discussed. These enhancements include use of temporal information in masking for audio segments with sudden change of amplitudes in them, efficient quantization of the variances and run-length coding for the zeros. Experiments suggest that these enhancements give almost transparent compression of monophonic CD quality audio signals (sampled at 44.1 kHz) at bit rates of 55-63 kbits/sec. We also discuss a real time implementation of a simplified version of the wavelet based audio coder at somewhat higher bit rates (64-78 kbits/sec).<>