Is There Mental Telepathy in Our Future?

June 17, 2015 – Researchers in Germany and the United States are decoding speech from brainwaves and turning them into text in a “brain-to-text” system that records signals from the cerebral cortex of epileptic patients. In an article entitled, “Brain-to-text: decoding spoken phrases from phone representations in the brain,” in Frontiers in Neuroscience, the researchers describe their study done at Albany Medical Center, Albany, New York.

The seven volunteer subjects had electrodes implanted in their frontal and temporal lobes of the cerebral cortex recording their brain activity during speech production. The technology used was an electrocorticographic grid or ECoG. ECoG was combined with waveform recording of the acoustic pattern of each subject’s speech. Voice data was recorded as well by microphone.

The test involved the volunteers reading text from a range of source materials scrolled across a screen at constant rates between 42 and 76 words per minute. Each exercise was repeated as often as three times. The researchers then segmented the text into smaller blocks which were then fed to a speech recognition engine to obtain specific acoustic phrase-based models aligned with ECoG waveforms.

In the final step researchers observed words and phrases in “unseen spoken utterances recognized solely by neural signal activity.” The error rate was as low as 25%, about one in four words. Now that kind of error rate is a little better than a coin flip but far from perfection.

In working with voice recognition systems I know I can get up to 97% accuracy converting spoken words to text but that takes lots of practice. For detail work that can mean a “not” disappearing from a sentence changing the meaning from a negative to a positive. And don’t get me started on homonyms. “Naught” and “not” are not the same and digital signal processors cannot discern the difference.

But this brain-to-text technology isn’t meant for someone like me. Think of it as the first opportunity for those who have been defined medically, as “locked-in,” those who cannot speak because they are in a vegetative state or comatose. For them communicating with a 25% error rate would be miraculous.

And if we can create technology that can read our unspoken thoughts converting them to text so that others know what we are thinking, we have moved one step closer to becoming telepathic.

Related articles across the web

• Speech recognition from brain activity
• Decoding spoken phrases from representations in the brain
• Scientists Pinpoint a Speech Center in Brain
• FDA Approves Cognivue Automated Cerebral Assessment System To Detect Signs of Dementia