That paper specifically mentions which electrode locations were used: "The sites of the left- and right-side scalp electrodes (locations C3 and C4 over sensorimotor cortex) that controlled the cursor are marked."
Which seems to miss the key C3 and C4 sensors which would allow mental visualizations of hand and feet movements for more refined Cognitiv detection. More recent work from the same research institute also incorporates the Cz sensor.
It feels like there is a degree of "uninformed guesswork" going on in the Emotiv forums around which mental models to use for training. If published papers like the above were available as early as 2004, why weren't those sensor locations included, perhaps substituted for some of those in the current array? It seems like they would have been obvious choices for the types of applications the Emotiv headset is designed around, specifically with regard to Cognitiv.
Can anyone comment on design decisions around the placement of electrodes found on the production EPOC headset?
Can anyone provide suggestions on research projects and published material which focus on the sensors the EPOC does provide?
Wow. This question deserves a lengthy and properly reasoned answer. I'll try to prepare one, but the key part of the answer is that we found no advantage to including those locations as actual sensors. C3 and C4 are also pretty tough to get to with a lightweight, simple wrap-around headset such as we offer. The purpose of the EPOC is to do the best possible BCI job with the least difficulty in fitting for the average consumer, who does not want to mess around with straps and adjustments - and these were found to be necessary to access the top-sides of the head on all comers. The mechanical challenge is too big for the payoff - C3 and C4 tend to push the headset off the top of large heads and are a bugger to fit to small heads. Simple as that!
Our initial proof-of-concept testing used a full neoprene medical grade headset and we captured data from 32 locations. We did some serious stats on this and we found no advantage for a group of over a dozen researchers in training and controlling Cognitiv actions using the existing set of sensors against larger subsets of the 32 - there was no statistical difference in performance with or without the sensorimotor cortices. Don't ask me why - we were making a product, not trying to fathom the murky depths of our researchers' brains. No measurable performance advantage, distinct mechanical disadvantages = go with what we have. Note that we could have removed more sensors for only a minor performance cost - but we chose not to for reasons of redundancy. If you have a funny shaped head or drying sensors the detections will continue to work until you have taken away too many sensors. We have patents pending on how we do this
The purpose of the EPOC is to do the best possible BCI job with the least difficulty in fitting for the average consumer, who does not want to mess around with straps and adjustments
That makes sense, and thanks - as always, for the quick response!
When more time permits, if yourself or anyone else visiting these forums can offer advice regards the second question, that would very much be appreciated as well:
Steve Castellotti wrote:
"Can anyone provide suggestions on research projects and published material which focus on the sensors the EPOC does provide?"
I think for my own work the next step will require further research directed along those lines. Much of the software I am interested in developing is geared towards the education sector, particularly with the younger set, and the biggest challenge will be providing simple mental models for training which are based on knowledge of what is and is not detectable by the hardware.
For example if C3 and C4 had been available and I knew from published mu rhythm research those areas were related to sensorimotor pathways for feet and left and right hands, it is easier to insist the users imagine moving those parts of their bodies, because the mental model is backed up by facts. Personally I am very motivated to work with this technology and therefore have a fair degree of patience for pushing through frustration. That is not often the case when working with early adolescents!
Anything which helps to minimize guesswork or even unique instruction on a per-user basis will contribute a long way towards a successful pilot project.
If your working with adolecents you may have difficulty fitting the headset. It may be possible to jury rig extensions to the sensors which would allow you to relocate them where you want, although this may necessitate the research edition as it may affect the normal processing of the cognitive suite. I'm just guessing on the second thought about cognitive.
Cognitiv is very non-picky about the sensor arrangement as long as (a) there is reasonable coverage of the brain and (b) the placement is fairly consistent between fittings, if you want to preserve saved signatures. Some visualisations will work a little better or a little worse with different sensor arrangements - but there are two parts to the system: the EPOC and the user. Both are generally quite efficient at learning and adapting to circumstances.
Fixed location algorithms like Expressiv and Affectiv won't take well to different locations, but Cognitiv (and trainable Expressiv, and whole-brain detections like Excitement) adapts very nicely to whatever is presented.
A bioinformatics company offering a unique platform for crowd-sourced brain research. It leverages cloud computing, big data and mobile technology to offer valuable personal insights and accelerate brain research globally.