Main Page
From HumanSense
Contents |
The Digital Rashomon Project
From Wikipedia: "The Rashomon effect is the effect of the subjectivity of perception on recollection, by which observers of an event are able to produce substantially different but equally plausible accounts of it."
The purpose of this project is to further the capabilities of mobile devices in perceiving the context of their users. Location-based services are beginning to gain traction, but the context that they extract is limited to physical location in absolute coordinates, and the services that they provide are, therefore, limited.
However, the sensing capabilities of modern smartphones present opportunities for extracting much richer contextual information and for modeling user behaviour at a much higher level.
As a simple example, most location-based services rely on absolute coordinates, making use of data from GPS or A-GPS. The coordinates are used to instantaneously place an individual on a map, and this information is used to present the user with opportunities, usually commercial, related to the location of the user. History is, at best, used to find a set of locations that a user has visited to build up a rough profile. However, simple fingerprinting techniques can be used to find locations that a user frequents, indoors or outdoors, and other techniques can be used to estimate modes of travel between locations. Here, we are not concerned with localization in terms of absolute coordinates on some map, but with identifying locations that are specific to individual users. We view locations as places where people spend time, and learn locations that are specific to individual users.
Together with temporal information, these data can be used to generate stories of a users' routines. If data is collected over a long enough period of time, rhythms and patterns can be harvested from this data that can describe behaviours. This project is concerned with developing methods for analyzing large amounts of contextual data and with creating models that can capture and explain both individual and group behaviours.
People
- Jordan Frank <jordan.frank@cs.mcgill.ca>
- Actively seeking interested parties. Email Jordan if interested.
Related Publications
- A Novel Similarity Measure for Time Series Data with Applications to Gait and Activity Recognition (UBICOMP 2010, adjunct proceedings), with S. Mannor and D. Precup, Sept. 2010.
- Activity and Gait Recognition with Time-Delay Embeddings (AAAI 2010), with S. Mannor and D. Precup, July 2010.
Platform
We are targeting the Android OS, and we are currently using the Google Android Developer Phone 1 (ADP1) and Nexus One devices.
Here is the main page for the Android Data Collection Software
Other Hardware
We have also tried tinkering with and building our own sensors, as well as some fantastic sensors that Intel was kind enough to provide. However, currently we are targeting off-the-shelf smartphones, and so this section is outdated.
We have done some work with the Nokia n810.
The original sensor boards that we considered were based on the Arduino bluetooth board.
Intel has graciously provided us with two Mobile Sensing Platforms (MSPs).
Software
- See the Google Android Developer Phone 1 page for instructions on using the Android software.
- VMWareInstructions for installing VMWare Player on the agent machines.
- Good Resource on using Bluetooth stack in Python.
- Data collection on the N810
Resources
Help on the Wiki
- Consult the User's Guide for information on using the wiki software.
- Configuration settings list
- MediaWiki FAQ
- MediaWiki release mailing list
