The research described in this thesis investigates the area of subband coding of high quality digital audio signals. The research into the application of a nonuniform filter bank employing frequency warping to digital audio coding is described. Firstly, some aspects of psychoacoustics, especially masking effects, are studied. Utilising psychoacoustic models, perceptual entropies are calculated for a number of audio excerpts and it is shown that theoretically the bit rate can be reduced to lower than 2 bits per sample. Different types of filter banks are reviewed. To examine the best band splitting scheme, studies are carried out on optimum bit allocations for uniform and nonuniform filter banks, with or without a psychoacoustic model. The concept of perceptual coding gain is proposed and it is shown experimentally that band splitting according to the auditory critical band rate is not recommended for audio compress10n. A new generic nonuniform decomposition algorithm, relying on cascading frequency warping and a uniform filter bank, is developed. Detailed analyses of frequency warped signals are carried out. The effect of truncation of the original signal and the warped signal is investigated. It is shown that the reconstruction is near perfect and coding based on the new nonuniform filter bank is capable of noise shaping. A subband coder based on the nonuniform filter bank is constructed and results are presented. Objective assessment and informal subjective assessment are carried out and they show that the coding reaches high quality at 192 kbps per channel; however, some artefact is perceptable. An efficient implementation of the Koilpillai-Vaidyanathan pseudo-quadrature mirror filter (KVPQMF) bank, which features simplicity in prototype filter design and suitability for audio coding, is developed. The savings in the number of calculations and in the execution time over the direct implementation is significant.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:320382 |
| Date | January 1996 |
| Creators | Xiang, Wei |
| Publisher | University of Central Lancashire |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://clok.uclan.ac.uk/18834/ |
Page generated in 0.0016 seconds