A recent paper published on presents an exciting new approach to decoding speech directly from non-invasive brain recordings. This could pave the way for restoring communication abilities in patients who have lost the capacity to speak due to neurological conditions. The study provides hope that with continued research, non-invasive brain decoding could give a voice to the voiceless. arXiv How did they figure out how to turn brain recordings into speech? Let's see.
Significant Improvements in Accuracy For 3-second segments of speech, the model could identify the matching segment from over 1,500 possibilities with: Up to 73% accuracy for MEG recordings Up to 19% accuracy for EEG recordings This represents a dramatic improvement over previous attempts at speech decoding using non-invasive sensors. It also approaches the accuracy achieved in studies using invasive brain implants.