Packet loss concealment for voice transmission over IP networks
Ejaz Mahfuz · 2001
Voice-over-IP (VoIP) uses packetized transmission of speech over the Internet (IP network). However, at the receiving end, packets are missing due to network delay, network congestion (jitter) and network errors. This packet loss degrades the quality of speech at the receiving end of a voice transmission system in an IP network. Since the voice transmission is a real-time process, the receiver cannot request for retransmission of the missing packets. Concealment algorithms, either transmitter or receiver based, are used to replace these lost packets. The packet loss concealment (PLC) techniques described in the standards ANSI TI.521 (Annex B) and ITU-T Rec. G.711 (Appendix I), have good performance, but these algorithms do not use subsequent packets for reconstruction. Furthermore, there are discontinuities between the reconstructed and the subsequent packets, especially at the transitions from voiced to unvoiced and phoneme to phoneme. The goal of this work is to develop an improved PLC algorithm, using the subsequent packet information when available. For this, we use the Time-Scale Modification (TSM) technique based on Waveform Similarity Over-Lap Add (WSOLA) to reconstruct the dropped or lost packets. The algorithm looks ahead for subsequent packets. If these packets are not available for reconstruction, algorithm uses information from past packets. Subjective tests show that the proposed method improves the reconstructed speech quality significantly.