Implementation of packet-based encoding schemes for speech transmission

Missing packets due to network congestion can seriously degrade the quality of received speech in voice packet communication. The natural redundancy of speech offers some protection against loss and about 1 to 2% missing packets cause little noticeable distortion if silence is substituted for the missing segments. The implementation of packet-based embedded coding (PBEC) and the performance of a simple encoder/decoder which exploits PBEC are discussed. The encoder/decoder combines the sample-by-sample predictive mode necessary for low bit rate operation with whole packet prediction for missing packet reconstruction.