A Multilanguage Study of the Quality of Interleaved MELP Voice Traffic Over a Lossy Network

The quality of time sensitive interactive voice traffic can seriously degrade if moved over lossy wired or wireless IP networks, particularly with low rate speech. One technique that can be used to improve the quality is to transmit interleaved, and possibly duplicated, voice frames. This paper compares the quality of non-interleaved clear MELP voice traffic with the quality of several interleaving schemes over a high loss packet network. Multiple NATO languages were tested. The results show that interleaving can significantly improve the perceived voice quality, but at the expense of an increase in mouth-to-ear delivery delay and, if redundancy is enabled, an increase in required bandwidth. The improved voice quality can be quite significant at high packet loss rates.