Scientists Test a Brain Computer Interface that can Decode Speech up to 62 wpm

In this modern age, research and development have reached new levels. We get to see new technological breakthroughs and innovations every other day. Simultaneously, in a recent breakthrough, a group of Stanford researchers tested a brain-computer interface (BCI) that can decipher speech at up to 62 words per minute, which is 3.4 times faster than the previous record. It would be a significant step toward real-time speech conversion that corresponds to the speed of casual human conversation.

The Stanford scientists discovered that in order to translate brain activity into coherent speech using a machine learning algorithm, they only needed to examine brain activity in a very small part of the cortex. The objective was to restore the voice of those who are unable to speak due to ALS or a stroke. While some degree of communication has been restored for people with paralysis thanks to keyboard-based solutions, decoding might be substantially sped up with a brain-based voice interface.

How the experiment was conducted?

This is the first time a BCI has significantly outperformed the communication rates that alternative technologies, like eye tracking, can offer for people with paralysis. A person with ALS who can move his mouth but has trouble producing words participated in an experiment in which the scientists recorded neural activity from two tiny regions of their brain. The researchers then converted these signals into words at a startlingly quick rate using a recurrent neural network decoder that can forecast text.

In spite of paralysis and limited cortical surface coverage, they discovered that monitoring these orofacial movements and the brain activity that goes along with them was “potentially robust enough to enable a speech BCI. However, the system wasn’t flawless. The recurrent neural network (RNN) decoder developed by the researchers still had a 20% error rate.

In an email, Max Hodak, who co-founded the BCI startup Neuralink with Elon Musk but was not engaged in the study, referred to it as “a substantial step change in the utility of implanted BCIs.”

Check out? IBCC has approved the equivalence of the CP to the HSSC science group.