Decoding algorithm based on MMSE for speech codec

An error resilience scheme for very low bit rate speech coding was developed for speech communications on narrow band channels with high bit error rates. The minimum mean square error (MMSE) is computed for each decoding state to estimate optimal parameters and to reduce the influence of the bit error. The codec states were estimated by computing the maximum posterior transition probabilities with a simplified computing method. The proposed scheme requires less computations and is less complex than previous methods. Objective and subjective simulations show that the performance of the proposed scheme is always better than that of the frame erasure method at all bit error rates.