Adaptive Wireless Multimedia Services

Abstract: Context-awareness is a hot topic in mobile computing currently. A lot of importance is being attached to facilitating the user of various mobile computing devices to provide services that are more “user-centric”. One aspect of context-awareness is to perceive variations in available resources, and to make decisions based on the feedback to enable applications to automatically adapt to the current environment. For Voice over IP (VoIP) software phones (softphones), variations in network performance lead to fluctuations in the quality of the communication. Therefore, by making these softphones more adaptive to the network environment will, to some extent, mask such fluctuations. Dynamic voice and video adaptation derives from the fact that different coder-decoders (CODEC) have different characteristics, even the same CODECs with a different configuration can behave quite differently, in terms of bandwidth consumption, packet size, etc. Minisip is a VoIP client application which was implemented on and targeted for a Linux platform. One of my tasks was to port Minisip to Microsoft’s Windows Mobile operating system, running on an HP IPAQ Pocket PC H5550. Such handheld computer enables the user to communication while they are moving about, thus increasing the probability that the characteristics of the network connection will change. Building upon this port, the next task was to add dynamic voice and video CODEC adaptation. Dynamic voice and video CODEC adaptation on Minisip poses several challenges, for example, in what way can the network performance be determined and what adaptation strategy can achieve high call quality while making efficient utilization of available network resources. In order to make the proper design choices, several estimation models will be discussed, these are used to determine an efficient, un-intrusive, and light weight means of dynamic CODEC selection within Minisip. This thesis only implemented audio CODEC adaptation of Minisip, and the evaluation of the resulting prototype shows that such dynamic adaptation is both feasible and practical; further more, video CODEC adaptation would be a more significant extension to this work in the future.