3 ms·

I've spent a lot of time working with EEG and pulling my hair out over garbage SNR, finnicky setups, and artifacts galore. I don't understand why so many folks

by kortex 9y ago

I've spent a lot of time working with EEG and pulling my hair out over garbage SNR, finnicky setups, and artifacts galore. I don't understand why so many folks try to go the route of using EEG as mouse/keyboard/generic HID for the masses, when (ignoring para/quadraplegics) most humans have no problem with computer input. There are tons of applications where the EEG can be used for what it's (sorta) good at: attentional and mood states, which, in the time of distraction firehose, seems like it would have a use case. But no, we have to make yet another pick-up-a-virtual-object interface. (not to imply X-aplegics are not worthy of tech. it's just that as mentioned above, eye tracking is WAY more robust)

hprotagonist 9y ago
Yup. IN fact, our application was locked-in patients. SSVEP BCI "works", but only if you can foveate on the target, which means you can move your eyes, which means you should use an eye tracker.
- yoz-y 9y ago
  There are quite a lot of papers that show you can actually modulate SSVEP response even without eye movement. Their protocol was quite fascinating, they used a swarm of targets with four frequencies all intermingled. You had vertical and horizontal (short) lines which were either blue or red. One would control the system by concentrating on blue horizontal stripes for example. Of course the throughput was quite low even for a BCI. I think more research in this vein could reveal some interesting possibilities.
  - hprotagonist 9y ago
    You can. In our experience, we were unable to get reliable performance with SSVEP or m-sequences with targets in the periphery, though - the subject had to be foveating on the target to get a good response. Given that constraint, an eye tracker would be just as applicable to the patient population, so SSVEP was a bag of disadvantages with no obvious upside.
beekunrah 9y ago
Yes the primary commercial application found in the EEG literature seems to be fatigue monitoring for professional drivers. Certainly "control" BCI applications won't be the first Neurotech killer app.
Sir_Substance 9y ago
>pulling my hair out over garbage SNR I bought into the OCZ NIA about a decade ago. Nice unit, but I was using it within about 2m of a set of AC/DC plugpack transformers. It had a SNR visualiser, and you could see it go increasingly ballistic as I moved my head towards and away from the powerboard. I have my sincere doubts that EEG devices can ever function satisfactorily in residential homes. Too many microwaves and fridges, turns out there's a reason hospital EEGs are done in faraday cage rooms.

by kortex 9y ago

hprotagonist 9y ago

Yup. IN fact, our application was locked-in patients. SSVEP BCI "works", but only if you can foveate on the target, which means you can move your eyes, which means you should use an eye tracker.

yoz-y 9y ago
There are quite a lot of papers that show you can actually modulate SSVEP response even without eye movement. Their protocol was quite fascinating, they used a swarm of targets with four frequencies all intermingled. You had vertical and horizontal (short) lines which were either blue or red. One would control the system by concentrating on blue horizontal stripes for example. Of course the throughput was quite low even for a BCI. I think more research in this vein could reveal some interesting possibilities.
- hprotagonist 9y ago
  You can. In our experience, we were unable to get reliable performance with SSVEP or m-sequences with targets in the periphery, though - the subject had to be foveating on the target to get a good response. Given that constraint, an eye tracker would be just as applicable to the patient population, so SSVEP was a bag of disadvantages with no obvious upside.

beekunrah 9y ago

Yes the primary commercial application found in the EEG literature seems to be fatigue monitoring for professional drivers. Certainly "control" BCI applications won't be the first Neurotech killer app.

Sir_Substance 9y ago

>pulling my hair out over garbage SNR I bought into the OCZ NIA about a decade ago. Nice unit, but I was using it within about 2m of a set of AC/DC plugpack transformers. It had a SNR visualiser, and you could see it go increasingly ballistic as I moved my head towards and away from the powerboard. I have my sincere doubts that EEG devices can ever function satisfactorily in residential homes. Too many microwaves and fridges, turns out there's a reason hospital EEGs are done in faraday cage rooms.