0.8 kb/s high quality speech coding algorithm
Song Du · Journal of Tsinghua University(Science and Technology) · 2003
This paper presents a 0.8 kb/s high quality vocoder to obtain high quality synthetic speech at a very low bit rate. The algorithm is based on a traditional linear prediction model with multiframe joint vector quantification, multiband mixing excitation, subband voicing strength parameter prediction, adaptive spectral enhancement, and pulse dispersion etc. Simulation results demonstrate that the intelligibility of this 0.8 kb/s vocoder is quite good and the natural tone is fine. The diagnostic rhyme test (DRT) score is 85% in an informal test. Furthermore, the vocoder is robust in a noisy environment, and intelligible even with a bit error rate of 10-2.