February 19, 2015 at 3:02 pm #105641
Reassembled the epoc headset for a better fit.
February 19, 2015 at 3:02 pm #219966
Reconstructed my epoc to be a bit more sturdier, and focuses all of the sensors in one target area.
Since it uses elastic, it fits snugly to your head and keeps the contacts secure to prevent potential artifacts
from head movement.
I used a cardboard backing and poked holes in it. Then i glued a thin layer of foam to it for comfort.
the hard part was feeding the contact wires through the cardboard and foam, but i needled my way through.
After securing the wires to the contacts, i glued the contact sensor housings to the foam padding,
Next, i spread a thin layer of glue over the foam padding, to prevent the foam from being soaked
with saline/water. (this was to prevent mold and potential shorts if it ever made it to the circuit board)
The cardboard was backed with a thin layer of aluminum foil for its EM blocking potential. I used
electrical tape to create a nice housing and sleek look for the unit.
The battery and both circuit boards are loacated behind the aluminum ‘shielding’ and secured with
The elastic band is secured to the sides of the cardboard for a tight fit.
the reference sensors are located on the sides, i figured this might be crucial to the workings of the unit.
The unit works well and acquiring the quality signals takes less time as the elastic ensures a tight fit.
However, as gmac has stated before in previous posts, having the sensors so close together in close
proximity isn’t going to give you much difference in EEG readings, as the different contacts are
collecting similar bursts of neural activity.
The nice part of this unit is that you can rotate it around your head to measure the desired areas,
although granted some positions will be slightly more comfortable than others. (ie using it as a chin
strap takes some getting used to. lol)
If I could have designed it over again, i may have spread out the contacts a bit more.
tell me what you think, i welcome your criticism and suggestions.
February 20, 2015 at 5:40 pm #219968
Interesting! Guess I don’t need to mention the warranty is kinda voided 😀 facial expressions and emotional detections are also going to be meaningless. I would be interested to see if you can get good results with mental commands, they are pretty resilient. You will need to place the sensor patch very close to the same location if you want to re-use your training profile. Gyros will still work but the axes may be a little messed up depending where you mounted the control board a d where you fit the device.
Close proximity sensors like that are fine for detailed imaging of specific brain regions – this is about the density of a 256-sensor array. Shame you lose all the diversity of sensors spread around the head, but depends what you want to do! Good luck, let us know how it goes
February 23, 2015 at 11:30 am #219989
lol yes i was going to make the warranty joke myself but can’t have all the fun 😀
yeah there is a lot of compromises doing it this way, and you’re right about losing a lot of the diversity.
you’re making me wish i had left it alone. lol but i’m going to stick with it and see what its capable of.
i never really used the facial expression feature as much anyways, it kept making the sad face. lol
i’m going to start out by subtracting all sensor locations from each other, since most of the sensors will
be taking samples of similar neural readings. i figure this will deduce the variance from the small area
i’m reading from. though it will surely be small differences. however, big things have small beginnings.
not too worried about the gyros either, although in theory they should work just as well, the way its mounted
i’d probably have to invert X/Y.
i was actually more worried about the reference point sensors, being so close together to the sensors and
even relatively closer to each other, i wasn’t sure how it would affect it adversely, but everything seems to
be well in synch.
February 23, 2015 at 7:13 pm #219991
May e look at some of the tricks you can do with high density headsets. The literature is full of Laplacian processing and source localisation which gets better local resolution with high densities. I would also consider moving the references away a bit – at least CMS,
February 28, 2015 at 10:44 pm #220049
i did some reading as you suggested on Laplacian algorithms, proved to be an interesting read.
sounds like the technique received a slow start from the EEG community as it wasn’t a well received theory.
sounds like its best for precise signal confirmation.
also learned about convolution matrices and how they work, i discovered its used frequently in image processing as well
and discovered my graphics editing software is capable of performing this feature. i should find this feature quite useful
for cleaning up image details. its funny how things can be right in front of you, and you never notice them until you
change your perspective.
i may port this algorithm for unity, though it really sounds like a job for matlab and i can guarantee it would harbor such
a feature. ill keep you appraised of what i end up with, thanks for the idea!
You must be logged in to reply to this topic.