Ayumu: Efficient Lifelogging with Focused Tasks
stodds at cs.duke.edu
||Monday, November 21, 2016
||10:00am - 12:00pm
||North 311, Duke
Today's lifelogging devices capture images periodically without considering what data is important to users. Due to their small form factors and limited battery capacities, these lifeloggers are bound to miss important data either because they record at a slow rate to conserve power, or because they record at such a high rate that they must frequently recharge. In this paper, we present a new approach to lifelogging that better utilizes a device's battery by integrating knowledge of the specific information that a user wants captured. We have developed the first instance of such a focused-task lifelogging system called Ayumu, which aims to capture the reading material that a user interacts with over the course of a day. Instead of capturing images periodically, Ayumu uses a suite of inexpensive sensors to record only when reading material is present. By recognizing when it would be most beneficial to capture images, Ayumu can achieve superior precision and comparable recall to a conventional, periodic lifelogger while using less energy.
Advisor(s): Ashwin Machanavajjhala and Landon Cox
Committee: Jun Yang, Benjamin Lee