Finite-state VQ excitations for CELP coding of speech
Adil Benyassine, H. Abut · 2002
The authors propose an improvement to the excitation model of CELP (code excited linear prediction) coders by embedding a labeled-state finite-state vector quantization (FSVQ) in constructing 5 ms long excitation vectors. Experiments indicate that the quality of the synthesized speech is very similar to those of the full search CELP systems requiring 1000 b/s more.>