A novel speech coding approach is proposed that gives high-quality performance at 8 kb/s. The coder is based on an autoregressive moving-average (ARMA) model as the prediction filter and a new excitation model. A simple ARMA analysis method is proposed that features a technique for eliminating the fine harmonic structure within the speech spectrum. This overcomes the problem of MA parameter misestimation. The excitation signal is modeled as a pulse train whose density is varied depending on the residual signal's power. The proposed coder produced high-quality speech comparable with 6-b log PCM at 8 kb/s according to computer simulation.
|Number of pages
|Proceedings - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing
|Published - 1989
|1989 International Conference on Acoustics, Speech, and Signal Processing - Glasgow, Scotland
Duration: 1989 May 23 → 1989 May 26