Understanding human tasks from wearable sensor signals

The Human Centric AI Seminars Series – DSS Group

The Human Centric Security team are running a new monthly series “The Human Centric AI Seminars” that will focus on various research topics in human centered AI.

For more info contact: Kristen Moore and Tina Wu. Free access to anyone interested in Humans and AI.

Prof. Julien Epps

Abstract: A ‘task’ is arguably the most basic unit of human activity and perhaps the most logical basis from which computing systems can understand humans, yet at present we have only extremely limited means by which to detect a change in task and to estimate the level of physical, mental and other types of load experienced during tasks. Recent activity in wearable computing promises change, with the opportunity to position non-invasive near-field sensors directly where they are most useful for task analysis, for example in front of the eye, near the mouth and fixed to the head. Head-mounted wearable sensors allow ‘always-on’ automatic analysis of tasks and behaviour based on these sensor signals, and hence automatic approaches to the long-studied but historically manually-intensive problems of workload assessment and task switching analysis. This seminar introduces a framework for automatic task analysis and highlights a number of findings from automatic processing of eye activity, speech and head movement that advance the state of the art towards realising it. Finally, some methods for automatically detecting and modelling low-level behaviour indicators in a longitudinal manner will be outlined, which pave the way towards a big data understanding of human behaviour.