Quality of service for VoIP in wireless communications

• Main Author: Lopetegui Cincunegui, Iban
• Published: University of Newcastle Upon Tyne 2011
• Subjects: 621.382
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.545769

Ever since telephone services were available to the public, technologies have evolved to more efficient methods of handling phone calls. Originally circuit switched networks were a breakthrough for voice services, but today most technologies have adopted packet switched networks, improving efficiency at a cost of Quality of Service (QoS). A good example of packet switched network is the Internet, a resource created to handle data over an Internet Protocol (IP) that can handle voice services, known as the Voice over the Internet Protocol (VoIP). The combination of wireless networks and free VoIP services is very popular, however its limitations in security and network overload are still a handicap for most practical applications. This thesis investigates network performance under VoIP sessions. The aim is to compare the performance of a variety of audio codecs that diminishes the impact of VoIP in the network. Therefore the contribution of this research is twofold: To study and analyse the extension of speech quality predictors by a new speech quality model to accurately estimate whether the network can handle a VoIP session or not and to implement a new application of network coding for VoIP to increase throughput. The analysis and study of speech quality predictors is based on the mathematical model developed by the E-model. A case study of an embedded Session Initiation Protocol (SIP) proxy, merged with a Media Gateway that bridges mobile networks to wired networks has been developed to understand its effects on QoS. Experimental speech quality measurements under wired and wireless scenarios were compared with the mathematical speech predictor resulting in an extended mathematical solution of the E-model. A new speech quality model for cascaded networks was designed and implemented out of this research. Provided that each channel is modelled by a Markov Chain packet loss model the methodology can predict expected speech quality and inform the QoS manager to take action. From a data rate perspective a VoIP session has a very specific characteristic; exchanged data between two end nodes is often symmetrical. This opens up a new opportunity for centralised VoIP sessions where network coding techniques can be applied to increase throughput performance at the channel. An application layer has been implemented based on network coding, fully compatible with existing protocols and successfully achieves the network capacity.