PACKET LOSS CONCEALMENT USING AUDIO MORPHING

In packet switched networks (case of IP networks) no minimal threshold on rate is offered for the transport of packets. The available bandwidth depends on conditions of the traffic on the network (best effort strategy). In VoIP (Voice Over Internet Protocol) audio packets must be continuously played. When packets are lost or have arrived too late it gives place to chopped sound and to a degradation of the quality. In this paper the lost case is considered when a packet loss is recognized by the reception of the next packet. Generally concealment techniques do not use this received signal. We propose here an algorithm, which uses this information to improve the missing signal synthesis and allows maintaining an acceptable quality of the speech signal. A comparison between this method and techniques recommended by the ITU-T for coders G.711 [3] and G.723.1 [1] will also be described. 1