VoIP Voice and Fax Signal Processing

5.1 PACKET LOSS CONCEALMENT OVERVIEW

VoIP voice is based on real-time. Internet Protocol (IP) packet delivery. Reducing end-to-end delay for IP packets delivery is important for voice quality. Hence, any loss of packet cannot wait for retransmission. Packet impediments are caused by several factors, namely quality-of-service (QoS) actions on the sending side, network impediments, drops in downstream, jitter buffers, and so on. Details on IP packet impediments are given in Chapter 10. These net results of packet drop can result in nonavailability of voice frame for voice decoding. A PLC operation details how to synthesize voice during a resultant packet drop.

Voice codecs based on ITU G.711, G.726, G.729AB, and G.723.1A are popularly used in VoIP applications. The vocal-tract-based codecs G.729AB and G.723.1A have built-in packet loss concealment operations and operate on a vocal tract model for packet loss concealment. These codec models have a long memory, and, hence, it takes more time to recover from lost packets. G.711 and G.726 are waveform-based codecs, which recover quickly from loss conditions to correct speech. G.711 is a sample-based codec that can recover the packet loss state to the correct state instantaneously. G.726 maintains a short memory of previous conditions, but it recovers more quickly than G.729 and G.723.1. Because of relatively quick transitions in G.726, the techniques used for G.711 are applicable for G.726. In this chapter, PLC techniques are described ...