Tom Lovett and Eamonn O’Neill Department of Computer Science University of Bath Bath BA2 7AY UK +44 (0) Social sensing: context transitions
What? Detecting the occurrence of a context change, e.g. location or activity change Why? Improving user self-reporting tools, notification delivery, bootstrapping context- aware systems How? Inferring context from motion sensing on a mobile device
Context transitions: challenges What constitutes a ‘significant’ transition? What are the limitations of a mobile device? –Power can limit sensor sample frequency –CPU (and power) can limit ‘online’ local processing What sensors/sensor combinations are good indicators of a transition? Can we detect transitions without expensive processing?
Context transitions: benefits User self-reporting tools –Improve on current systems that use ‘random beeping’ or rely on user remembering to report Bootstrapping –Lightweight detection can trigger context dependent processes Context driven notifications and services –Beyond a research tool
Context transitions: how Mobile device motion sensor fusion (beyond the accelerometer) Binary yes/no – has a transition occurred? Not what has occurred Tuning parameters, e.g. sensor weightings, to capture significant transitions and ignore the insignificant Tradeoffs: power vs accuracy; spam vs information loss
Issues The challenge of “social context” –e.g. several meetings in the same place (same activity, same location, different social context) Are virtual sensors better social sensors? –e.g. users calendars, social networks How may we legitimately sense social data in a privacy conscious world?
Thank you