Honda Robot Controlled By Brain Waves

Dotnaught writes "Honda researchers to have developed a way to control robots using human brain waves. Using brain signals read from a person in a magnetic resonance imaging scanner, a robotic hand mirrored the movement of the human controller, spreading its fingers and making a 'V' sign."

Comments

I've always thought...

[Temujin_12]

...that it was only a matter of time until we started to see brain-to-machine mappings for communication. The possibilities are very exciting (coding with your brain anyone?). What scares me is when efforts are taken to have machine-to-brain communication. Call me crazy, but I prefer my own synapses to be the only source of thought in my brain. I don't even want to begin to think what it could happen when the machine segfaults (or gets hacked into) while injecting thoughts into my brain.

universality?

[Red Flayer]

So, does the MRI interpreting algorithm need to be tailored to each user, or could an 'off-the-shelf' interpreter work for anyone?

While I'm sure that bloodflow signatures for physical movements are similar between individuals, is there too much variability to prevent false recognition of a 'signal'?

Any neurobiologists out there care to help out?

Re:universality?

[venicebeach]

So, does the MRI interpreting algorithm need to be tailored to each user, or could an 'off-the-shelf' interpreter work for anyone? While I'm sure that bloodflow signatures for physical movements are similar between individuals, is there too much variability to prevent false recognition of a 'signal'?

At this point, it is surprising that they can even do it for an individual (discerning among these three quite similar hand movements). I am kind of skeptical myself. There is a lot of variability in fMRI signal even within an individual. I would guess the system is trained on a specific individual.

Between individuals you have additional sources of variability; for example, the foldings of the cortex are quite different from person to person. I personally have a very unusual precentral gyrus on the left side. Activity maps are typically aligned to anatomical maps so finding correspondences between individuals has to deal with the challenges of anatomical variability.

For gross things, it can be quite obvious what the person is doing. I can tell by looking at the activations in your brain if you are looking at something versus hearing something. But looking at a duck versus looking at a cow? Much harder. Making a V-sign versus making a fist? I've never seen a paper where someone reported being able to do this. It is theoretically possible, but difficult with a blurry MRI signal that aggregates over populations of neurons. You can certainly do it if you implant electrodes into the brain. Recordings from monkey premotor cortex, for example, find neurons that fire when specific movements are made.

like the monkey stuff from a few years ago

[kaan]

About 3 years ago, some scientists hooked up a chimpanzee and captured brain signals to control a robotic arm. Their results were quite a bit more impressive I think, because the robotic arm had full motion control, and was physically located several hundred miles away from the chimp. But still, this stuff from Honda is cool, because it's controlled by humans using mri, not wires plugged into your brain like the monkey stuff. I just hope they don't try to put brain controller stuff in their vehicles...

Here's an article from New Scientist:
http://www.newscientist.com/article.ns?id=dn4262