Algorithms for high fidelity audio compression
Talal G. Shamoon · 1996
Algorithms for encoding high fidelity audio at low bit rate are presented. The ultimate goal of this work is to provide fast algorithms for sub-band/perceptual coding that are capable of audio coding at bit rates that permit transmission of high fidelity digitized audio over broadcast and telephony channels, and storage of digitized audio on low capacity media. In order to achieve this, the standard paradigm for sub-band audio coding is extended. A new filter bank that provides non-uniform band partitioning that more closely resembles that in the human auditory system is presented. Further, the use of adaptive differential pulse code modulation quantization is explored. The principle contribution of this thesis is a rapidly adaptive lossless compressor that can be used in a sub-band coder. This algorithm maps variable length strings from the source alphabet onto variable length output strings, and achieves a performance boost over standard compression algorithms in use. An entropy estimation algorithm and a source model for digitized audio are also presented in this vein. While these techniques have been applied to audio in this thesis, they can easily be adapted to video and speech sources.