Degree Name

Master of Engineering (Hons.)


Department of Electrical and Computer Engineering


Prototype Waveform Coding is one of the most promising methods for speech coding at low bit rates over telecommunications networks. This thesis investigates quantisation mechanisms in Multi-Prototype Waveform (MPW) coding, and two prototype waveform quantisation algorithms for speech coding at bit rates of 2.4kb/s are proposed. Speech coders based on these algorithms have been found to be capable of producing coded speech with equivalent perceptual quality to that generated by the US 1016 Federal Standard CELP-4.8kb/s algorithm. The two proposed prototype waveform quantisation algorithms are based on Prototype Waveform Interpolation (PWI). The first algorithm is in an open loop architecture (Open Loop Quantisation). In this algorithm, the speech residual is represented as a series of prototype waveforms (PWs). The PWs are extracted in both voiced and unvoiced speech, time aligned and quantised and, at the receiver, the excitation is reconstructed by smooth interpolation between them. For low bit rate coding, the PW is decomposed into a slowly evolving waveform (SEW) and a rapidly evolving waveform (REW). The SEW is coded using vector quantisation on both magnitude and phase spectra. The SEW codebook search is based on the best matching of the SEW and the SEW codebook vector. The REW phase spectra is not quantised, but it is recovered using Gaussian noise. The REW magnitude spectra, on the other hand, can be either quantised with a certain update rate or only derived according to SEW behaviours.