BSN Contest Description
For this contest, organizers will make multiple limited data sets, accompanied by appropriate meta data, available on this website. Each data set is collected using in-house tools by one of the contributing groups. Based on these limited data sets, registered teams will design signal processing algorithms to perform specified transformations, classifications, and event detections. An algorithm for a particular function (e.g. stride time variation in human gait inertial data sets) can be parameterized based on the provided meta data, but it must otherwise be unaltered when applied to the different data sets.
In particular, each data set contains of two groups of
movements. The first group is composed of transitional movements including 'sit
to stand', 'sit to lie', etc. The second group includes walking data on a treadmill at
various speeds, with different inclinations, etc.
The contest will specifically ask to implement signal
processing algorithms, for the first group, to classify movements and
approximately determine their timing information (i.e. the beginning and the end
of each known movement with a precision of 500 ms).
The objective for the signal processing on the second group,
which consists of walking data, is to determine the stride time.
The details of sensor specifications and placement are posted under the data set section.
On the day of the contest, organizers will make extended data sets available to the contestants. These data sets will be in the same formats as the original limited data sets, but the data may have been collected from different human subjects and have some variations in data collection protocols. Teams will then be evaluated on how well the algorithms, developed based on the limited data sets, perform on the extended data sets.
We would like to acknowledge Telecom Italia for their generous sponsorship of this contest and the prizes.
- $1,000 cash prize and a contest winning certificate will be awarded to the fisrt place.
- $500 cash prize and a contest winning certificate will be awarded to the first runner-up.
- The competition welcomes participants from academia and industry. (In future years, K-12 institutions will also be invited to participate.)
- Registration must be complete before the registration deadline (firm deadline). No late registrations will be accepted.
- The signal processing can be implemented in any language.
- The implementation needs to be stand alone. During the competition, the code may have to execute on the referee's computer, which may not include all of the software/libraries used in development. See FAQ for more information.
- The signal processing algorithms must not rely on any network communications.
- It is expected that the participants will not have prior experience with software and tools developed and data collected in the organizers' laboratories. After the registration, the organizers will carefully evaluate the list of participants, and reserve the right to identify any conflict of interest. Any conflict of interest or prior involvement with organizers' laboratories will disqualify the team to compete in the contest.
- The contest is limited to 20 teams. Participants are selected on a first come first served basis, so please register early.
- Accuracy - how well the signal processing performs the specified function compared to the ground truth -- See FAQ for details.
Day of Contest Logistics
- We will release the test data sets. Each team will be given two hours to process the data and send us the results. You will be required to send us both your results and your code (executables or source code -- up to you!), with enough instruction so we can execute them.
- We have teams competing from all over the world. Hence, it was challenging to select a two hour period that would be convenient for everyone. Therefore, we decided to adopt the following approach: Email us (firstname.lastname@example.org and email@example.com) and let us know a convenient time beginning Friday May 20th until Sunday May 22nd (around noon CDT - Dallas time). We will send you the link and the password to the data sets on the time that you wish to have it, and you will have two hours to email us your results and the code. If you think email cannot handle your code, we can arrange an alternative. But it is important that you send us your results in two hours. Your code should produce the exact same results that you are sending us.
- We will ask you to fill out a brief survey after the competition. Please give us as much feedback as possible.
Visit our Frequently Asked Questions section. If you have further questions, contact Roozbeh Jafari (firstname.lastname@example.org) or
Louis Atallah (email@example.com).