Excellent Video Tom, I think ill have to invest in an anti-static mat. I see you were using an extension cable for the dongle. Can inteference be reduced by moving the dongle away from the PC by this method or does all the interference come from the headset? I also have a wireless router right in front of my PC, I may have to try relocating this. Oh and is Electra free this Saturday night

Do a Google search on "High Density Black Conductive Polyethylene Foam" or "anti-static foam" I was think of creating a head with this material.

After reviewing the forum posting there no Standardization, another reason to developed a standard head for testing maybe the <B> the anti-static foam head </B> is the answer.

Cheers

It will be less if you use real biosensors. You are measuring an electrolyte-filled balloon of brains, bones, muscles, blood vessels and evil thoughts. Your test probes have a "contact potential" which is nonlinear with current flow and results in significant errors in impedance. Biosensors tend to contact through an electrolyte using a non-polarisable contact material (where the electrolytic potential on the material is independent of the current flowing through the interface). I wonder where you might find some of those babies?
In any event, the relevant measures are impedance (not just resistance) and capacitance (for electrostatic testing especially - yeah, I know, to all those pedants out there - capacitance appears in the impedance hehehe - I was pointing out the charge storage characteristic is quite important. A small cauliflower will react with bigger voltage diversions than a large human, although perhaps more attractively in my case).
Working on some other suggestions and comments Tom!

Thanks Dan,
I am glad someone appreciates some of the problems I am trying to deal with. Based on the electronic testing you have done, do you think this head set is capable of producing repeatable brain maps if the outside interferences such as static are controlled?

Quote
Tom Martin wrote: ..do you think this head set is capable of producing repeatable brain maps if the outside interferences such as static are controlled?..

Well Tom, I've just finished reading ~24 research papers listed on the EEGLab website http://sccn.ucsd.edu/eeglab/ and two books on EEG, several times, and all I can say is maybe. The task is proving that what you end up with is actually brain waves. For example, this paper published in 2009: "Validation of regression-based myogenic correction techniques.." references a study that demonstrates, using facial muscle paralysis, that the majority of EEG above 25 Hz is actually EMG. The paper also describes methods for separating EEG and EMG data at lower frequencies. It was a wake up call.

Long shot

How about "Static Guard Anti-Static Spray" sold at some drug stories

Hi Tom,
I'm not sure what level of immunity you are looking for (fairly amazing, I guess - you're already at the level of wired medical EEG systems), but you might try wearing the headset and putting a tinfoil helmet over the top. If you ground the foil helmet to your head somewhere (tuck it under your glasses, or under the rubber pads) you will find amazing quietness. Melons, cauliflowers and any other vegetables or fruits will not give a viable representation of your expected signals - the capacitances and connections are just not really close to humans.
Good luck - the other advantage is that aliens can't implant thoughts or read your mind (except those from the planet Tinfoil)
Best regards,
Geoff

The headset is not made from true anti-static plastic. The conductivity of those materials attenuates the EEG signals unfortunately. The construction uses polycarbonate/ABS alloy. Arms are currently constructed from polypropylene but will change to polycarbonate/polyester alloy next build run.

What is the reason for requiring this level of anti-static performance guys? Most electrostatic effects are transient and tend to get swamped in any fft measurement, especially over long integration periods. As I said, medical grade EEG suffers from electrostatic problems as well. I'm concerned you are putting a lot of effort into a losing battle which doesn't make a huge difference anyway. I'm speaking from experience with developing our own detections. These clearly work pretty well without the kinds of precautions you are describing, although it's true you get the occasional false trigger from electrostatic events and it would be nice to avoid them. But you will never get rid of the problem until we change the humans and the environment - ask any EEG researcher anywhere about this. Tinfoil hats really help if you need a special setup. We already use state-of-the-art methods to combat electrostatics (we use ultra-fast transient absorbers on each input channel, we use standard CMS/DRL circuits etc etc) - the only thing we don't have is a ground wire, which helps but does not cure the problem.

Thanks a lot tom for the videos...they give really valuable insight into the current technology....

please post more videos if possible using the tinfoil method as geoff suggested.

Are'nt there any brainwave patterns that are more distinctive than the others.?? for example...doesnt shouting or placing a finger near your forehead produce a distinct brainwave pattern.

Furthermore rather than trying to detect patterns in the data visually ,have you considered more complex pattern recognition approaches such as artificial neural nets etc.

Thanks
Sumeru

Quote
Tom Martin wrote: Could the gold contacts under the black "comfort pads" be used for grounding?

The headset is connected to the head through the CMS connection (the left-hand reference) and the DRL connection (the right-hand reference). The gold plates you refer to are wired in parallel with CMS and DRL. Use these points a reference connections instead of the default locations (or even as well as, if you have spare sensors) to further reduce noise. They hit skin directly and there's another bit of common-mode noise eliminated.

If you hook an earth wire to this point you will certainly have an impact, but it may not be what you hope for... it's all part of a complicated feedback loop, so who knows what can happen. By all means try it

Quote
Tom Martin wrote: Some think I am being unreasonable to Emotiv in its formative years if I just point out vulnerabilities.

Tom, we have no problem with criticism, especially when it's based on fact and measurement, as you are doing. Things need to be in proportion though - even the world's most accurate atomic clock has errors, but for calculating whether you're going to miss your bus, it's a little bit of an overkill to try to refine your clock - you're better off figuring out how to spend less time in the bathroom. (Sorry, extreme example and I don't want to minimise your effort - I applaud it and welcome the dialogue).

In our defence, (a) we have taken all reasonable and many unnecessary steps to reduce electrostatic interference, especially for the type of product and intended use (including neurofeedback), (b) we suffer almost the same level of interference as the best available medical devices - of course they have the luxury of waiting for clean air and the ability to control the measuring environment, which is not an option for us, and © the spectral characteristics of the interference are usually way out of the range of interest for most things - for example you would need to stand close to a rotating fan with just the right speed to trip an alpha-wave detection. That is, a 4-blade fan would have to be spinning at 150 rpm to show up - more like a propellor than a fan...

The bottom line is, these artefacts are real and unavoidable - we are dealing with humans, electronics and real environments. The real challenge is to deal with them, and only when absolutely necessary should you need go to extreme lengths to reduce the effect. As I said, our detections are quite robust because we took into account real-world effects from the start. Sometimes they go off by themselves - but is that really electrostatic interference or poor signature definition? Blinks, winks and eye movements, and to a lesser extent Excitement, are more susceptible than the other detections to electrostatic events. We designed this thing to be worn by real people bouncing around their lounge rooms. It's pretty damn good at that, but not perfect - nothing ever is, in engineering and science.

Quote
Tom Martin wrote: I think the aluminum foil cap may have real promise, but I haven't tried it yet. Perhaps we can have an Epoc competition for the most creative design. Does anyone know Origami? A propeller on top would show a lack of understanding of the Triboelectric affects of moving air.

Love the idea. Even the propellor - see above

Quote
Tom Martin wrote: Shouting or moving your finger near the headset is exactly what you don't want to do , as moving air and moving hair from shouting produces a lot of static artifact. Moving a hand that could contain a static charge near the head will also affect things inconstantly. Tying long hair back in a pony tail/bun after the headset makes contact greatly reduces the hair static.

Tom, honestly I don't think it's fair to advocate people don't move or act naturally while wearing the headset. We have managed to demonstrate through fairly diligent experiments and extensive user experience at demonstrations and shows that the EPOC is easily robust enough to handle these events. Reading this thread, you would think you have to lock yourself in a humidicrib wearing grounded metal mesh clothing to get the headset to work.

It's great that you have found some things you can do to improve your experience with the headset - but the difference overall between doing nothing and taking extreme precautions is at most a couple of percent in overall accuracy. It's just not that big an issue under most circumstances! Electrostatic events occur - but we've done a hell of a lot to minimise their impact. Our detections are pretty immune to them - you would never have seen this without looking in detail at the signals, and it's really easy to convince yourself that the jumps and bumps in the signal are a bad problem - they really aren't because we (and every other EEG researcher) work around them.

Quote
Tom Martin wrote: I wish Control Panel and future BCI programs would have a "Correction Dialog Box" like the voice recognition programs do. That way if you thought Pull and it Pushed you could select a Correction Box to choose the proper action to associate that thought with. I think this would speed and improve the training considerably.

This is a good suggestion, but hard to implement in practice - especially in gameplay. Speech recognition is not really a perfect analogue - you know how to speak and you've known how to speak for many many years. The stimulus is very consistent. With brain signals, you've never done this before. At least half the time, the Correction Button will need to be on the human, rather than the machine. It's a mutual learning experience (possibly the speech recognition analogy is better than I thought - you CAN learn to speak so the damn thing understands you, it just isn't picking up natural speech).

What we have is a little more laborious, but if you're triggering a false action: (1) add more data to the intended action, (2) add more NEUTRAL data - can;t have too much , and then if things are still a problem, (3) add more data to the falsely triggering action. It turned out this was the best and fastest way for us in house and with volunteers, because of the mutual part of the experience (human learns to repeat distinct patterns + machine learns to recognise repeated pattern, while still remembering slightly confused data from earlier).

Tom, I'd like to thank you for your contribution to this. I applaud your efforts and I'm glad to see users are taking thse things seriously. Please don't be offended with a little push-back. We really did engineer this device as well or better than anything else I've ever made - medical diagnostic and therapeutic products included - and it is easily good enough for its intended purpose. We are equally obsessive about quality and good design and we pushed the envelope a long way with much of this. We agree things can be improved and we love to see others thinking about extreme uses and how to get there. Let's not minimise what it's good at - but equally let's be honest and open about what you have to do if you want to achieve more difficult tasks with the equipment.

If people want to work with raw EEG signals they should (a) work out what level of electrostatic noise (and other noise) can be tolerated by the intended use, (b) learn what can be done to reduce the electrostatic noise to the necessary levels, and © try to develop means to deal with residual noise in the experimental analysis (it may be a simple as tagging events and excluding data during that period, or as complex as building neural nets and predictive algorithms to extract underlying signals from events as they occur). Finding good answers at © may well feed back into part (a)

It is starting to look like I am over reaching not only the present day capabilities of the Epoc, but of EEG science in general. Very sophisticated computational models and algorithms seem to be needed to make any sense at all of this electrical data we are looking at.

Hojjat Adeli the Editor-in-Chief of the International Journal of Neural Systems has a brand new book "Automated EEG-Based Diagnosis of Neurological Disorders: Inventing the Future of Neurology" that describes integrating neural networks, wavelets, and chaos theory to identify epilepsy (probably the easiest condition to identify) with 96% confidence.



I am obviously naive thinking I can identify which one of eight possible neurological conditions is present by looking at brain maps generated with a an Epoc or any other EEG for that matter.

Just the same I will still keep trying to find a "Strange Attractor" in all this chaos.