A High Quality0.6Kb/s Speech Coding Algorithm Based on MELP
Bin Qing-yuan · Audio Engineering · 2004
This paper presents a0.6Kb/s vocoder based on MELP.In the vocoder three consecutive MELP frames are grouped into a superframe and jointly quantized utilizing the interframe correlation properties.High qual-ity synthesized speech is achieved.The algorithm uses such techniques as two-frame LSF joint quantization and bi-directional prediction in LSF quantization,distinct pitch quantization schemes for different U/V modes in a su-perframe,statistical VQ codebook design in bandpass U/V decisions quantization,mean-removed vector quantization for gain parameters and a better gain interpolation algorithm in the decoder and etc.