Adaptive source and channel re-encoding to maximize tandem link performance

We investigate the issues in voice communications over tandem heterogeneous network connections, which require multiple source decoding/encoding operations. While it is known from speech coding research that tandem connections of different speech coders can extract a substantial performance penalty, it is shown that the use of error concealment and source re-encoding can improve system performance in the presence of channel impairments such as bit errors and packet losses. A new protocol is proposed for wireless LANs that reduces retransmissions and discarded packets, and hence reduces system delay and facilitates packet loss concealment.