A method for comparing the performance of EVS and other voice codecs under bursty packet loss

Abstract

The use of IP technology for delivering voice services is growing rapidly across all type of networks (fixed and wireless). One of the impairment factors that IP networks exhibit which affects the quality of voice services is packet loss. Different types of networks will have different packet loss characteristics, with most being susceptible to bursty loss, which can be significantly detrimental to voice quality. As part of the evolution of codec technologies, Enhanced Voice Services (EVS) has been introduced by 3GPP in order to provide higher speech and audio qualities at lower bitrates and higher resilience to packet loss compared with previous communication codecs. In this paper, a method has been developed for characterising and comparing the performance of codecs under different bursty packet loss characteristics. It has been used to compare the performance of EVS with AMR-WB and G.722. EVS has been tested with and without its partial redundancy mode, known as Channel-Aware (CA) mode. All codecs have been tested using a range of different packet loss profiles (PLPs) with loss rate values reaching 20% and mean burst length values going up to 10 packets. It was found that EVS with and without CA performs better than AMR-WB and G.722 for almost all the tested PLPs.