Neural network based event estimation on lifelog from various sensors

The data related to our life experiences is called lifelog, which can easily be collected with mobile electronic devices in recent years. Although lifelog research has been conducted for a long time, practical applications such as a memory assistant system have not been fully developed yet. This is mainly due to the lack of methods to structurize the lifelog data efficiently. In our research, we developed a method for structuring a lifelog consisting of data from various sensors, focusing on event estimation with neural network. In an evaluation experiment, we captured lifelog data with a device that has various sensors, and then we estimated the events, i.e., the participantsf activities. As a result, the system correctly estimated events 70.4% of the time. We also created a lifelog viewer to visualized the data based on the result of event estimation.