The CELP decoder primarily consists of an excitation source and a synthesis filter. Additionally CELP decoders
often also include a postfilter. The excitation source has both periodic components, contributed by the adaptive
codebook, and random components contributed by one or more fixed codebooks. At the decoder, the excitation
signal is reconstructed using the codebook indices (pitch lag for the adaptive codebook and shape index for the
fixed codebook) and gain indices (adaptive and fixed codebook gains). This excitation signal is then filtered by
the linear predictive synthesis filter (LP synthesis filter). This filter is obtained by interpolating the LPC
coefficients of successive analysis frames, where the LPC coefficients are reconstructed using the LPC indices.
Finally, a postfilter is applied in order to enhance the speech sound quality.