GSM uses a Regular Pulse
Excited (RPE) codec. Input is split into 20ms frames, each with 8 short term predictor coeffiecients. Frames are further split into four 5ms sub-frames, each with a delay and gain for a long term predictor.
After short and long term filtering, the residual signal for each sub-frame is decimated into three possible excitation sequences, each 13 samples long. The sequence with the highest energy is chosen as the best rep-resentation of the excitation sequence, and each pulse in the sequence has its amplitude quantized with three bits.
At the decoder the reconstructed excitation signal is fed through the long term and short term synthesis filters to reconstruct the speech. A postfilter improves perceptual quality