ELSC Seminar Series

Abstract

Loss of speech due to injury or disease is devastating, and severely affects the
ability to communicate with other people. Therefore, paralyzed persons unable to
speak may benefit tremendously from neuroprostheses for speech restoration. In this
talk, I will present our novel speech neuroprosthesis that artificially articulates
building blocks of speech for a silent person based on single-neuron and high
frequency activity in brain areas never harnessed for a neuroprosthesis before.
However, until recently, single neurons in these areas were not characterized as
speech-related. I will therefore describe some of our underlying research which
uncovered the neuronal populations in multiple areas of the human brain and the
encoding they utilize to code for speech features.
Our studies also employed these neuronal populations, from the deeper structures of
the brain to the cortex, for decoding single neuron activity to infer speech features.
The contribution of our research is therefore triple. From the basic science point-of-
view, it expands our understanding of the human speech control system and
elaborates our knowledge of the physiology of several brain areas. From the
methodological side, we developed AI-based decoding algorithms and a speech
neuroprosthesis, and tested them both offline and online. Finally, from the clinical
viewpoint, our study bears an enormous potential for completely-paralyzed patients
to be able to “speak” again with other people, at will.