A team of Stanford scientists claims to have tested a new brain-computer interface (BCI) that can decode speech at up to 62 words per minute, beating the previous record by 3.4 times.
That would be a huge step toward real-time speech conversion at the pace of natural human conversation.
Max Hodak, who co-founded BCI company Neuralink with Elon Musk but was not involved in the study, called the research “a meaningful step change in the utility of implanted BCIs” in an email to Futurism.
As described in another peer-reviewed paper, the team of Stanford scientists found that they only needed to analyze brain activity in a relatively small portion of the cortex to convert it into coherent speech using a machine learning algorithm.
The goal was to give back those who can no longer speak or caress their voice due to ALS. While keyboard-based solutions have allowed people with paralysis to communicate again to some extent, a brain-based voice interface could significantly speed up decoding.
“Here we have demonstrated a speech BCI that can decode unlimited sentences from a large vocabulary at a rate of 62 words per minute, the first time a BCI has tracked the communication speeds that alternative technologies can provide for people with paralysis, e.g.”, write the researchers.
In an experiment, the team recorded the neural activity of an ALS patient, who can move his mouth but has difficulty forming words, from two small areas in his brain.
Using a recurrent neural network decoder that can predict text, the researchers then converted these signals into words — and at a surprisingly rapid rate.
They found that analyzing these orofacial movements and associated neural activity was “probably strong enough to support a speech BCI, despite paralysis and limited coverage of the cortical surface,” the paper said.
But the system was not perfect. The researchers’ recurrent neural network (RNN) decoder error rate was still about 20 percent.
“Our demonstration is a proof of concept that decoding attempted speech movements from intracortical recordings is a promising approach, but it is not yet a complete, clinically viable system,” the researchers admitted in their paper.
To improve the error rate of their system, the scientists propose to examine more parts of the brain while optimizing the algorithm.
More about BCIs: Neuralink co-founder says new company’s eye implant could be the “ultimate” VR technology