Improved lost frame recovery techniques for ITU-T G.723.1 speech coding system

Accepted by the ITU in 1996, the G.723.1 dual rate speech coder [1] has since been used for application in low bit-rate videophone systems [2]. Specifically, it has been used to encode voice for multimedia services over packet-switching networks including IP, ATM, Frame Relay, as well as over mobile communications networks. Due to the near toll quality [3] of the G.723.1 speech coder, it is ideal for real-time voice communications over private and local area networks (LANs) where packet loss is minimal. However, over wide area networks (WANs), global area networks (GANs), and mobile communications networks, congestion is often severe and increasing rates of packet loss may generate significant distortions in output speech quality if unremedied. This paper describes a lost frame recovery technique for reconstructing lost packets at the speech decoding system. For packet loss rates up to 15%, simulation results show a substantial improvement in output speech quality over the current strategy [1] provided for the G.723.1 speech coder, at the price of increasing algorithmic delay.