Researchers Use Artificial Intelligence to Convert Brain Activity Into Speech

Researchers from Columbia University, University of Bremen in Germany, and Maastricht University in the Netherlands have reportedly made headway in converting electrode data from the brain into computer generated speech with the aid of computational models known as neural networks.

 

The researchers said that it was difficult to infer the speech sounds from the neuron activity patterns as it is not straightforward and varies from person to person. The can also only implant the electrodes on an open skull in a limited time frame during surgery, and such an invasive recording process can only be done in rare cases

.  

Nevertheless, the researchers were able to get the neural network to map electrode readouts to the audio recordings, which then reconstructed words from previously unseen brain data. This yielded up to 40 percent of understandable computer-generated words.

 

Source: Science Mag