Research On Phase Information In Waveform Interpolation Speech Coding

Recently, low bit-rate speech coding has developed greatly. The research on low bit-rate coder at 4kb/s and below with communication quality has become one of the most attention at present. Waveform Interpolation as a great potential speech coder has got much attention.In traditional low bit-rate speech coding, considering that ears are not sensitive to phase information, the phase information is often neglected, and this will result in coarse and harsh speech quality, and it even may lead to inflection in pitch. In order to obtain a high-quality speech codec, the phase information of speech should be included in codec. In this thesis, a method for quantizing the phase of SEW (Slowly Evolving Waveform) and reconstructing SEW's phase with cubic polynomial interpolation is given based on the perceptual weighting analysis-by-synthesis (A-b-S) vector quantizer for the phase spectrum in WI coder. The subjective A/B listening tests indicate that the reconstructed speech's quality of this scheme is better than that of fixed phase and Complex Cepstrum with 4-6 bits. Moreover, a predictive phase vector quantizer is proposed in this thesis based on the above method. The synthesis speech is slightly improved for female speakers.In addition, an improved synthesis scheme of WI coder is presented in this thesis. In this scheme, the speech residual signal is synthesized directly using magnitude spectrum and phase track where there are continuous changes for pitch between frames, while the speech residual signal is synthesized using phase intergradations with a burst of the pitch. The computing complexity of WI decoder is reduced greatly by this method, and the reconstructed speech quality keeps invariable meanwhile.