Assessing the Quality of Audio and Video Components in Desktop Multimedia Conferencing

This thesis seeks to address the HCI (Human-Computer Interaction) research problem of how to establish the level of audio and video quality that end users require to successfully perform tasks via networked desktop videoconferencing. There are currently no established HCI methods of assessing the perceived quality of audio and video delivered in desktop videoconferencing. The transport of real-time speech and video information across new digital networks causes novel and different degradations, problems and issues to those common in the traditional telecommunications areas (telephone and television). Traditional assessment methods involve the use of very short test samples, are traditionally conducted outside a task-based environment, and focus on whether a degradation is noticed or not. But these methods cannot help establish what audio-visual quality is required by users to perform tasks successfully with the minimum of user cost, in interactive conferencing environments. This thesis addresses this research gap by investigating and developing a battery of assessment methods for networked videoconferencing, suitable for use in both field trials and laboratory-based studies. The development and use of these new methods helps identify the most critical variables (and levels of these variables) that affect perceived quality, and means by which network designers and HCI practitioners can address these problems are suggested. The output of the thesis therefore contributes both methodological (i.e. new rating scales and data-gathering methods) and substantive (i.e. explicit knowledge about quality requirements for certain tasks) knowledge to the HCI and networking research communities on the subjective quality requirements of real-time interaction in networked videoconferencing environments. Exploratory research is carried out through an interleaved series of field trials and controlled studies, advancing substantive and methodological knowledge in an incremental fashion. Initial studies use the ITU-recommended assessment methods, but these are found to be unsuitable for assessing networked speech and video quality for a number of reasons. Therefore later studies investigate and establish a novel polar rating scale, which can be used both as a static rating scale and as a dynamic continuous slider. These and further developments of the methods in future labbased and real conferencing environments will enable subjective quality requirements and guidelines for different videoconferencing tasks to be established.