A Mobile Dual VoIP System for Enhancing Speech Quality and Intelligibility: Simulation and Test bed

Abstract

As it is well known in a 3G/4G network scenario, the quality of voice traffic over IP (VoIP) is greatly reduced due to the strong current limitations in terms of requirements regarding delay, jitter, packet loss rate and guaranteed bandwidth. The present work highlights the benefits in terms of improved intelligibility when making a duplication of VoIP packets through two wireless data accesses provided by different operators. In particular, the paper presents the architecture and the prototype of a dual stream approach to mobile VoIP applications (Dual VoIP) over HSPA access networks. Test results, obtained via simulations and a real-time implementation of Dual VoIP algorithm, demonstrate an average packet loss reduction up to 90% and an average improvement of speech quality up to 1 PESQ point. Furthermore, the paper highlights the significant reduction of the audio signal clipping at all levels: phoneme, word, sentence and conversation. Enhancement of the speech quality and intelligibility of the audio signal is a very important aspect in common best effort applications using VoIP as well as in particular conditions such as environmental wiretapping for forensic uses and/or private tactical communications in network-centric contexts where real time listening and intelligibility of the speech signal play a key role. The deep evaluation presented here has the aim of understanding the behavior of the proposed architecture under different application scenarios and drawing, at the same time, useful conclusions on the improvement of the Dual VoIP prototype.