According to the MIT Technology Review, San Francisco scientists say that they’ve been able to create a device that can read brain signals to produce complete phrases. To do so, a group of University of California, San Francisco, researchers led by neurosurgeon Edward Chang first asked five people with epilepsy to speak from a list of 100 phrases. As they spoke, Chang’s team recorded their brains and brain signals.
The researchers then inputted the brain signal data into a computer program modeled after the human vocal system, and it was able to generate intelligible speech. Instead of trying to read our intentions or thoughts, the researchers’ technique monitored nerve firing patterns that controlled each subject’s vocal organs. A similar technique has been employed previously to allow brains to control robotic arms.
“We are tapping into the parts of the brain that control these movements—we are trying to decode movements, rather than speech directly,” Chang said.
Brain signals were documented through a pliable pad comprised of electrodes known as an electrocorticography array, or ECoG, and rested on the surface of the brain.
“This is probably the best work being done in BCI [brain-computer interfaces] right now,” said a University of Pittsburgh researcher, Andrew Schwartz. Listeners to the computer-generated speech were able to understand 50 to 70 percent of the phrases and words. According to Schwartz, if researchers managed to put probes inside brain tissue and not just surrounding the brain, the computer-generated speech would be much more accurate.