Back to bookshelf

An Instantaneous Voice-Synthesis Neuroprosthesis | Wairagkar et al. 

science•research paper•View Source
Appreciation
--
Importance
--
Date Added
8.28.26
TLDR
55b may encode what is being said rather than how it is produced. When T15 added intonation, activity shifted in v6v, M1, and d6v, but not in 55b.
2 Cents
--
Tags
  • T15 used the synthesized voice to ask questions or emphasize one word per sentence.
  • The authors split each array’s activity into output-potent dimensions that drove synthesized sound and output-null dimensions interpreted as preparation.
  • On the marked word, the output-potent/output-null ratio shifted in v6v, M1, and d6v at p≤10−11p \le 10^{-11}, but not in 55b at p=0.3p = 0.3.
  • Because intonation requires vocal-apparatus control, the shifting arrays may sit closer to execution while 55b represents content.
  • Caveat: the null/potent split depends on the fitted decoder, so 55b’s flat ratio may only mean acoustics map onto it less linearly. This is also one null result from one participant.