A robust transformed binary vector excited coder with embedded error-correction coding

Described a recently proposed approach for representing the excitation sequence in analysis-by-synthesis speech coders, where the excitation vector is given by the transformation of a binary vector with values -1 or 1. In this approach, the storage requirement and the exhaustive search of the excitation codebook are completely ruled out, resulting into an extremely simple encoding algorithm. The issue of embedding error-correcting coding into the speech coder to combat channel errors is also discussed. Both convolutional codes and Reed-Solomon codes are investigated aiming at the overall bit rate of 11.4 kb/s