Bandwidth efficient AMR operation for VoIP

An example of a bandwidth efficient adaptive multi rate (AMR) system for Voice over IP (VoIP) is presented. In VoIP, packet losses cause degradation of the synthesized speech. The distortions may propagate over several consecutive frames, since predictors in the codec exploit inter-frame correlations to gain coding efficiency. To reduce the effects of packet loss, forward error correction (FEC) that adds redundant information to voice packets can be used. However, while FEC can reduce the effects of packet loss, it will increase the amount of bandwidth used by the voice stream, which is not desirable. In this paper we propose FEC methods like partial redundancy, selective redundancy for the most sensitive frames and parameter interpolation in conjunction with AMR codec mode adaptation, which secure the speech quality when using AMR for VoIP without increasing the bandwidth substantially.