Decoding speech perception from single cell activity in humans

Ori Ossmy, Itzhak Fried, Roy Mukamel

Research output: Contribution to journalArticlepeer-review

Abstract

Deciphering the content of continuous speech is a challenging task performed daily by the human brain. Here, we tested whether activity of single cells in auditory cortex could be used to support such a task. We recorded neural activity from auditory cortex of two neurosurgical patients while presented with a short video segment containing speech. Population spiking activity (~. 20 cells per patient) allowed detection of word onset and decoding the identity of perceived words with significantly high accuracy levels. Oscillation phase of local field potentials (8-12. Hz) also allowed decoding word identity although with lower accuracy levels. Our results provide evidence that the spiking activity of a relatively small population of cells in human primary auditory cortex contains significant information for classification of words in ongoing speech. Given previous evidence for overlapping neural representation during speech perception and production, this may have implications for developing brain-machine interfaces for patients with deficits in speech production.

Original languageEnglish
Pages (from-to)151-159
Number of pages9
JournalNeuroImage
Volume117
DOIs
StatePublished - 5 Aug 2015

Keywords

  • Decoding
  • Local field potentials
  • Single unit recording
  • Speech perception

All Science Journal Classification (ASJC) codes

  • Neurology
  • Cognitive Neuroscience

Fingerprint

Dive into the research topics of 'Decoding speech perception from single cell activity in humans'. Together they form a unique fingerprint.

Cite this