Brain-Machine Interface Powered by EEG
From the wisbusiness.com article "Researchers use brain interface to post to Twitter"
Now, judging by the video, the way it works is that an inference algorithm constantly analyzes EEG measurements while letters are highlighted in random order on the screen. I'm guessing the highlighting events correlate with certain measurement spikes and to speed things up, the alphabet is organized into a grid where entire columns or rows can be flashed at once (faster than flashing the entire alphabet one step at a time -at least- once in order to determine whether a specific letter is being focused).
The approach is crafty but it also shows how crude current EEG technology is. If we had finer spacial resolution (by several orders of magnitude), we could probably do away with the on-screen alphabet entirely, instead relying on brainwave analysis alone which would require some previous training for both the brain and the machine, but after establishing a common "protocol" understanding would be pretty much automatic.
In early April, Adam Wilson posted a status update on the social networking Web site Twitter - just by thinking about it.
Just 23 characters long, his message, "using EEG to send tweet," demonstrates a natural, manageable way in which "locked-in" patients can couple brain-computer interface technologies with modern communication tools.
A University of Wisconsin-Madison biomedical engineering doctoral student, Wilson is among a growing group of researchers worldwide who aim to perfect a communication system for users whose bodies do not work, but whose brains function normally. Among those are people who have amyotrophic lateral sclerosis (ALS), brain-stem stroke or high spinal cord injury.
Now, judging by the video, the way it works is that an inference algorithm constantly analyzes EEG measurements while letters are highlighted in random order on the screen. I'm guessing the highlighting events correlate with certain measurement spikes and to speed things up, the alphabet is organized into a grid where entire columns or rows can be flashed at once (faster than flashing the entire alphabet one step at a time -at least- once in order to determine whether a specific letter is being focused).
The approach is crafty but it also shows how crude current EEG technology is. If we had finer spacial resolution (by several orders of magnitude), we could probably do away with the on-screen alphabet entirely, instead relying on brainwave analysis alone which would require some previous training for both the brain and the machine, but after establishing a common "protocol" understanding would be pretty much automatic.


