US Federal Standard MELP vocoder tactical performance enhancement via MAP error correction

The United States government has developed a new Federal Standard 2400 bps vocoding algorithm called MELP-mixed excitation linear prediction. This vocoder has a very acceptable voice quality under benign error-free channel conditions. However, when subjected to high error conditions as could be experienced in tactical vehicular operations, amelioration techniques may be employed which take advantage of the underlying inter-frame residual redundancy of the MELP parameters themselves. This paper describes experiments conducted on the MELP vocoding algorithm in conjunction with Viterbi convolutional error decoding, and enhanced with maximum a posteriori techniques which utilize these redundancy statistics. Both hard and soft Viterbi decoding implementations are investigated in addition to turbo codes.