Home
I recently completed my Ph.D as part of the Clarity Centre at Dublin City University. My PhD work was concerned with the utilisation of BCI (Brain Computer Interfaces) for image search and annotation tasks. I was supervised by Prof. Alan Smeaton and my work was sponsored by SFI through Clarity. More specifically I am interested in building image search systems driven by neurophysiological signals (ERPs) from users in repsonse to viewing images (P300).
I've done lots of other things straying away from my research area involving developing sensors for tennis practice through to developing hardware/software for virtual audio environments. There's many things I've done for fun and to expand my skillset, and other things I've done to narrow down my PhD direction. The thing is that I am interested in all things technology and science. Even though I will specialise, these broad interests will never change.
The basic crux of my research is working towards an image filtration/annotation system where:
The idea is that images which arouse your attention will be the images you have been told to look out for. The challange is telling by the brainwaves which images are the ones you are asked to search for. The images shown above are highly stimulating and contain lots of things that would catch our attention, making it very difficult to tell by brainwaves alone. If these images were less varied and you were for instance looking out for a brain tumour on a series of x-rays would it be easier to detect the relevant images by brainwaves? Is it easier if the target images are all visually alike? Is it easier if all the non-target images are alike? These are questions I am currently seeking to answer.
Why do this?
The idea is that you are able to sort large volumes of images much faster that you could ever do manually.
Why can't a computer automatically do this?
Computers are stupid. They only do what we can tell them to do. The computer cannot see the same
way that you or I can. Trying to describe to a computer what a cat looks like would be very difficult so that it could do it by itself, and even then the compupter
may not be very good at it.
We can however as people see things like cats easily, so why not use signals from our brain to help the computer in the process?
No Reason not to! We should be doing this!
Who would want to do this?
That's one question being looked into. Obviously people who need to look at and sort a lot of images. People at the ESA (European Space Agency) are interested as to whether their scientists would like to do this with specialist datasets where only a few of these scientists can really understand and pass judgements on the images (with their experience and expertise).
Maybe someday this will be used by people at home and in the workplace to help them sort images. Don't laugh! Think back to when somebody said this:
“There is No Reason Anyone Would Want a Computer in Their Home.“ - Ken Olson, president, chairman and founder of Digital Equipment Corp (DEC) in 1977 .
Is it worth all the hassle? Having all those wires connected looks annoying and not worth the effort.
That's one problem I am interested in. Maybe we do not need all these wires (channels). To study what we are doing we use lots of channels, but having used lots
of channels we can look at which ones are valuable and which ones are not. It might all depend on who is using this, and what they are searchign for but maybe you only need 4 or 5 channels to get near the same level of accuracy!
So what's the big problem?
There's lots of questions. Are some people better at detecting certain types of images based upon experience? Are some image sets and search tasks upon them better suited to this technique than others? We've shown that most people are able to detect certain types of targets easily, but does this generalise?