Erasure-robust speech coding and concealment in VoIP systems with frame packetization

In voice over IP (VoIP) networks, multiple voice frames can be sent within one packet to increase the network efficiency. When packet loss happens, the voice decoder often tries to conceal the erased frames from received parameters. This paper discuss the quantitative dependency of concealment quality on the packetization rate in terms of LPC distortion. The performance of an erasure-robust. concealment method is analyzed and simulated in comparison with the standard G.729 concealment method. Results show that around 1.2 dB improvement on spectral distortion can be obtained with erasure robust. concealment compared with G.729 under 1-6 frames/packet, packetization. A method to determine the maximum packet size given the network load and expected concealment quality is introduced. The proposed method serves as the packetization guidance during the call setup procedure in VoIP sessions.