Ref. [3],Figure 2
The decoder operates on a frame by frame basis. First the LPC coefficients are reconstructed out of the decoded LSP coefficients and therefore the synthesis filter can be rebuilt. Second for every subframe both the adaptive and the fixed codebook excitation are decoded and fed to the synthesis filter. The adaptive postfilter consists of a formant and a forward-backward pitch postfilter to improve the quality of synthesized speech. The quality is improvement is obtained by increasing the SNR at multiples of the pitch period. The excitation signal is input to the pitch postfilter, which in turn is input to the synthesis filter. Afterwards the output of the synthesis filter is input to the formant postfilter (also requiring the LPC coefficients). A gain scaling unit maintains the energy at the input level of the formant postfilter.