Mind Over Matter: Brain Control Interfaces Become a Reality_4

The Future ET: Can we expect the normal evolution of technology to reduce if not eliminate those performance limitations?
GS: That's a complicated question. Just making your headset smaller and with a more complicated package is not going to eliminate the subject-training requirement. Now if someone created a robust dry electrode that can operate well in a noisy environment, that would clearly make things better. That is something that future technology could provide. There are parts that can be mastered with technology, but there are always going to be parts that are difficult when you are recording signals from the scalp.
ET: Do you think we will be able to move away from primarily recognizing motor activity in the brain and start detecting patterns in areas like the visual and auditory regions? Also, what about those stories involving the use of fMRI machines to do just that kind of advanced detection?
GS: In regards to your second question, it is unlikely in my opinion that we will ever see an MRI machine small enough and portable enough to be worn on your head. So that makes those devices an unlikely candidate for a wearable BCI and will keep them confined to the laboratory. In response to your first question we are already using ECoG to detect what vowels and consonants you are speaking, and when you imagine speaking them. We've had several abstracts published on this research over the past year or so and we are going to submit the study we've done on this very soon. I think that soon we'll be able to detect what words you are speaking and imagining to speak. We're not there yet, but we are getting closer.
ET:How accurate is the system with vowels and consonants?
GS: Only 40 percent, and even that is when using a small set of choices consisting of either four vowels or four consonants and trying to pick one of them out of the set, which isn't all that spectacular especially when you compare that to a typical speech recognition system. However there are a lot of areas in the system that we think we can improve, and we have bright hopes for the future.
(Note: the claimed accuracy of several commercially available speech recognition systems is well into the high 90 percent range.)
ET:Is there anything else you'd like to say about the future of BCI?
GS: I believe that the future of BCI revolves around man and machine meeting each other midway, just like you and I do when talking. I may have a funny accent that you find hard to understand, but over time you learn to adapt to it and to understand me better. Just like with motor imagery, the solution lies with the optimization of both brain and the computer.