Interest-based life logging

Blum, M. Pentland, A. Troster, G. (2006), InSense: Interest-Based Life Logging, IEEE Multimedia, 13 (4), pp. 40- 48. The paper describes a wearable data collection device called InSense based on Vannevar Bush's Memex principles. allows users to continually collect their interactions as store them as a multimedia diary. It basically take into account the sensor readings from a camera, microphone, and accelerometers. The point is to "classify the users activities and "automatically collect multimedia clips when the user is in an “interesting” situation".

What is interesting is the types of categories they picked-up to develop their context-aware framework: they chose location, speech, posture, and activities—to represent many diverse aspects of a user’s context. They also have subcategories (for instance for location: office, home, outdoors, indoors, restaurant, car, street, shop)

The experience sampling approach works like that:

Subjects wear the system for several hours without interacting with it. Audio and acceleration signals are recorded continuously. The camera takes pictures once a minute and WiFi access points are logged to establish location. After the recording session, the user employs an offline annotation tool, which presents an image at a time, the corresponding sound clip, and a list of labels from which to chooseshowing sensor placement.

What is also curious is their description of their algorithm that calculates the current level of interest of an event based on the context classification. Why do I blog this? I am less interested in the purpose of the system itself (sharing material) but rather by the data extracted from context readings and how this could be used to tell a story (or to build up a narrative). Of course, given my interest in games, I see this device as intriguing and potentially relevant to map the first life experience with virtual worlds counterparts; it could go beyond current pedometer that control dogs.