An E-Model based adaptation algorithm for AMR voice calls

With the increase in femtocell deployments and the migration toward all-IP networks Adaptive Multi-Rate (AMR) voice calls are increasingly being transported over Internet Protocol (IP) networks. The default AMR adaptation mechanisms are designed to adapt to radio conditions between the terminal and basestation and do not consider backhaul networks. Although Explicit Congestion Notification (ECN) can now be used to allow adaptation to congestion in the backhaul, this requires that all network nodes be ECN capable which is not the case for many deployed IP networks. This paper proposes a terminal centric Quality of Service (QoS) adaptation mechanism for AMR that does not require any network support. The mechanism utilises a real time implementation of the E-Model to dynamically modify codec source encoding rates; a detailed description of utilising the E-Model in real time for voice quality assessment of AMR voice calls is provided. Furthermore, the implementation of an AMR voice call application for NS-3 is also provided. Results are presented showing the ability of the adaptation mechanism to match the current network conditions and achieve higher voice call quality.