Pitch Filtering in CELP Using a Time Scaling Approach

In this paper, a new approach to pitch filtering in the synthesis stage of a Code-Excited Linear Prediction (CELP) coder is described. With this method, time scaling/shifting of the original speech is performed on a block by block basis in order to provide a better match between the pitch pulses in the original and reconstructed speech. This technique allows for a re-duction of the pitch lag resolution, with the attendant bit rate reduction, of the CELP coder while maintain-ing the same speech quality. 1