Although speech can be decoded from human brain signals using brain-machine interfaces, progress is hampered by low decoding speed and poor accuracy. Here, the authors took advantage of the conceptual similarity between computerized language translation and translating speech-related neural activity detected by electrocorticogram. An encoder–decoder system was developed in which the sequences of signals obtained during spoken sentences were encoded into abstract representations and were then decoded into single words and assembled into sentences, resulting in language decoding at a speed similar to normal speech.
References
Original article
Makin, J. G. et al. Machine translation of cortical activity to text with an encoder–decoder framework. Nat. Neurosci. 23, 575–582 (2020)
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Lewis, S. Gained in translation. Nat Rev Neurosci 21, 300 (2020). https://doi.org/10.1038/s41583-020-0308-0
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41583-020-0308-0