A suite of open source and open hardware projects designed to detect and alert people to someone suffering from an epileptic seizure (fit).
This is a work in progress at the moment. The most promising version is based on a Pebble Smart Watch to measure movement of the arm during a tonic-clonic seizure. Other prototypes include:

• Pebble Smart Watch Detector
• Microsoft Kinect (depth camera) Detection
• Acceleration Detection
• Audio Detection
• Video Detection

See http://openseizuredetector.org.uk

Note that there are separate github repositories for the smart watch seizure detector app (https://github.com/OpenSeizureDetector/Pebble_SD), and the Android Companion App that raises alarms in response to detected seizures (https://github.com/OpenSeizureDetector/Android_Pebble_SD).

Those of you watching this repository for changes to the Pebble Seizure detector should watch the above repositories too, or you will not see updates.

It is based on an accelerometer monitoring movement. It uses a fourier transform to extract the frequency spectrum of the movement, and monitors movements in a given frequency band. The idea is that it will detect the rhythmic movements associated with a seizure, but not normal day to day activities.

If the acceleration within the given frequency band is more than a threshod value, it starts a timer. If the acceleration remains above the threshold for a given period, it issues a warning beep. If it remains above the threshold for a longer specified period, the unit alarms (continuous tone rather than beep).

This is a development version so it contains a real time clock and SD card to record the measured spectra to help optimise the device.

My initial intention is to mount this on a floor board on which our son sleeps (he will not sleep in a bed...). It may not be sensitive enough to pick up the movement through the floor, so it may have to be turned into a wearable device.

No working prototype yet....

Getting there. See http://nerdytoad.blogspot.co.uk/2013/03/first-go-at-video-based-epileptic.html.

This is the most promising looking option at the moment - the depth camera can eliminate background noise very well, leaving just the image of the subject to analyse. It also looks as though it is possible to detect breathing movements in the depth camera image. The thing that is missing at the moment is analysis of the image intensities to calculate rate of breathing. See http://nerdytoad.blogspot.co.uk/search/label/Kinect for examples of images etc.

Graham Jones, 03 January 2014. ([email protected])