The videos above show off a new speech synthesiser that converts vocal tract movements into intelligible speech.
This “articulatory-based” speech synthesiser, developed by a French team and unveiled in PLOS Computational Biology, uses deep learning algorithms. It could help build a brain-computer interface to restore speech for individuals with severe paralysis.
Speaking activates a region of the brain that controls the movement of the different organs (called articulators) of the vocal tract, including the tongue, lips and larynx.
Some people with language impairments still have this part of the brain preserved but signals between the brain and the muscles moving the articulators are interrupted.
Using deep learning algorithms, Florent Bocquelet from the French National Institute of Health and Medical Research and colleagues took these articulator movements and translated them into audible, comprehensible speech.
The study paves the way toward a brain-computer interface in which the synthesiser will be controlled directly from the brain.