Measuring Interactivity in Audio Teleconferencing using Conversation Analysis

The aim of this research is targeted directly at determining the quality of multi-party teleconferencing with or without the presence of spatialised audio and induced audio transmission delay. The objectives of the work are to explore, in an interdisciplinary manner, the combination of engineering metrics and work from the field of Conversation Analysis (CA) to fulfil this aim. Specifically, techniques are developed to identify conversation attributes using objective measures of speech performance to help determine the quality of experience and ease of conversational flow within a multi-party teleconference scenario. This thesis will also consider CA methods in order to explore further the interactional behaviour displayed during multi-party teleconferencing in relation to recognised group interactions. The assessment of the impact of transmission delay within a multi-party teleconferencing context is an important element of the work, along with developing task selection and design for conversation-based testing. This work has been achieved through the implementation of a series of conversation-based subjective tests which required over 100,000 data points to be manually reviewed. This work will also consider the design of a more effective subjective measurement/reference system than possible with current techniques, such as the mean opinion score (MOS) based subjective testing standard. The new testing methodologies are the first to show objective differences in multi-party speech interaction in the presence of differing delays and monophonic/spatialised presentations. In particular the work shows that higher transmission latencies have significantly different conversational parameters when compared to normal face-to-face meetings. However, when spatialisation is added, the effects of this latency are reduced. In conclusion this thesis demonstrates that CA can be used to objectively measure differences in multi-party teleconferencing with different degradations and presentation methods in ways that other, current techniques, cannot