Mark's Adventures in EEG

During the 2 days of exhibition Mastermind was used by about 30-40 people. most people gave good feedback on the the whole idea of controlling the computer with just their mind. Although the room was noisy most people were able to achieve a level of success that made them feel like they were really in control of the program. 

the main area of criticism was that the the feedback to tell them that they were being successful was not clear (i.e. that simply jumping from a mixed video to a clear video was not sufficient to show them they were having an effect). As a result I incorporated a couple of graphical queues that made it clearer when they were changing the display with “concentration”. This involved including a text message to “CONCENTRATE _ NEED MORE BRAIN POWER” when not concentrating and “CLEAR SIGNAL” when the concentration threshold was exceeded.

Further feedback suggested that it was not clear what to focus on. as a result I am developing a circle that changes with concentration level.

The noisy video (which users have no control of) at the start of the active app, also proved too long, so this was reduced to 8s from 16s, to allow users to get straight to the action.

After a number of days trying to get the program working with the returnJSON() method with little success I have made the decision to return to the long-hand version where the programming is written out within the draw() method. I will then launch the timing using the millis() function.

the rough outline should be like this

draw(){

connect to headset, filter signal and return JSON object containing EEG power data

if time < 30s {play the video}

if time >30s and < 45s {play “RELAX” message and record baseline EEG }

if time > 45s { launch the application

             if attentionActive = false {play the mixed videos}

              if attentionActive = true{play the clear video

} }

following on from our very productive meeting with Dr Marguritte Barry last week, I am developing a storyboard of what a user will actually experience when the enter the exhibit. this will allow us to focus on the final product that the user experiences and allow us to finally script the experience. 

in order the keep this app consistent with the other exhibits, and so that it gives the user the impression that they are interacting directly with a “mind reading machine” as opposed to simply a programme running on a computer screen, I have investigated the options of placing the sketch into a browser window. Current attempts focus on getting the sketch to display in a canvas element within a html page. 

Alternatively, it may be possible to simply display the sketch on the full screen. Superduper.og appears to provide an API which will allow the sketch to be displayed full screen. http://www.superduper.org/processing/fullscreen_api/

So I’ve finally got the app to a point where were can begin to test what actually constitutes “concentration”. As mentioned before there appears to be a correlation with increased concentration and increased power in the high frequency EEG bands.

To test this theory with the current EEG setup I created an experimental set-up with 8 participants. 

Step 1 involved getting the participant to relax while wearing the headset. Baseline EEG was then recorded for 1 min. They were then asked to watch a video on youtube where they had to pay attention https://www.youtube.com/watch?v=z9aUseqgCiY

The data was collected and the relative power of each powerband was analysed using a paired Student t-test. the data showed a significant increase in the gamma power bands when the participants were concentrating. 

Now that our Neurosky headset has arrived it is time to look at how to access, analyse and present the data. As mentioned elsewhere there is a well-developed python library looking at eeg analysis, so that was the first place to start. a number of repositories exist to access the Neurosky headset https://pypi.python.org/pypi/NeuroPy/0.1.

there also appears to be code available in javascript https://github.com/dluxemburg/node-neurosky,

however, when cosidering the overall product I decided to investigate the possibility of using Processing www.processing.org given the visual nature of the final product. there appears to be plently of code available online which might provide a solid basis to start coding.

the basic principle of EEG is that an electrical signal is recorded from the brain. By analysing this signal, it is possible to get insights into the workings of the brain. This area is well established in clinical research, but not so well studied in a commercial setting. CLinical EEG systems are extremely expensive and work by recording multiple channels of EEG from different areas of the brain. Some researchers have begun to develop open source code to help analyse this code.

One of these avenues is pyeeg, a python module to extract EEG information. I am currently investigating this as a means to analysing any EEG signals we record.