For the past 10 years Philip Kennedy has been implanting humans with electrodes, refining ways to record, preserve, and separate signals from the brain. Kennedy started Neural Signals, Atlanta, to work on brain-to-computer interfacing and is now aiming at restoring speech. Patients enrolled in the study, typically brain-stem stroke victims, are implanted with what's called a Neurotrophic Electrode. It is shaped like a cone with a hollow tip and four gold wires inside. “Brain tissue grows into the tip and we record electrical activity across the Teflon-insulated wires,” says Kennedy. The implant goes inside the brain tissue and is wirelessly powered by a power-induction system which lies just under the scalp.
The system works like this: “We have the patient say a phoneme in his head,” says Kennedy. Phonemes are the smallest unit of sound in a language, for example, the ‘b’ in book, or ‘th’ in that. “There's a different firing pattern for each phoneme. In one patient, the computer can recognize 32 of the 39 English phonemes.”
“In the training, we asked the subject to listen to a phoneme produced by a computer. That person would have to say the phoneme in their head. If the computer interpreted it correctly, the computer repeats the phoneme. Sometimes we hit 80% accuracy, sometimes 50% or less,” Kennedy notes.
And this is where the engineering challenges come in. For example, one patient in the study could produce some short phonemes like ma, da, and I,O, U, which is a big deal to researchers, but not much use to the patient. “The real challenge is to put the phonemes together as words,” Kennedy says. “Because phonemes can be produced as fast as twenty in a second, there's a lot of data to sort out.”
Kennedy and his team of mostly engineers are also working to get the signals to fire under the patient's control better than they are right now. “We're constantly improving the electronics,” he adds. They're also working on the power-induction system, getting more signals, and improving the electrode.
Kennedy's goal in the next year is to have the patient generate a hundred short words. “We want to resynthesize the phonemes to produce some intelligible language. It will be five to ten years to get the system producing conversational speech.”
Kennedy envisions this technology moving beyond medical applications. “There's potential for enhancing healthy humans by improving memory storage and calculating abilities ? almost as in the movie ‘The Matrix,’ where the main character uploads knowledge through a socket in the back of his head. I don't know if we'll be downloading exactly like the movie, but certainly this research could lead to something that today is unthinkable.”
Address: 5636 Lemon Ave.
Dallas TX 75209
Phone: +1 214 5203694