A Framework of Energy Efﬁcient Mobile Sensing for Automatic User State Recognition
MobiSys ’09 Proceedings of the 7th international conference on Mobile systems, applications, and services.
This paper introduces an Energy Efficient Mobile Sensing System (EEMSS) in order to collect the vast amount of information produced by mobile applications. The case is made that battery life will be spared if certain sensors are turned off when not needed. Three real-time conditions are checked to attempt to detect what state the user is in. They are: motion (coming from the acceleration), location (from the GPS), and background environment (from the microphone), and Wifi detection. Their prototype was evaluated on Nokia N95 devices with the following user states defined:
“Walking”, “Vehicle”, “Resting”, “Home talking”, “Home ent
ertaining”, “Working”,“Meeting”, “Oﬃce loud”,“Place quiet”,
“Place speech” and “Place loud”
The states and conditions for those states are described in a XML document. Once the condition(s) for one state is satisfied it moves on to the next and so on. The rate at which these sensors are polled have been manually assigned by the system designer. It was noted that manually setting these values is not optimal and a proposed algorithm was introduced that took into account how much power each sensor took. Based on the determined state only certain sensors will be pulled to achieve the most amount of context with the least amount of battery consumption.