Forward error correction for CELP encoded speech

Communication systems use low-rate vocoders to reduce the bandwidth requirements of transmitted speech. However, these algorithms introduce a new problem in that the quality of the reconstructed voice exhibits a much greater sensitivity to compressed parameter bit errors than it would to sampled speech bit errors. Furthermore, within each compressed packet, individual bits vary widely as to their contribution to overall speech quality. This paper analyzes the bit-error performance of CELP encoded speech and develops a forward error correction technique which maximizes speech quality using a given number of error correction bits.