BrainComputer NaturalLanguage Interface What is it An interface
Brain-Computer Natural-Language Interface ● ● What is it? An interface between the brain and a computer that enables natural language communication within the brain to be translated to external format a computer can understand. Ie, think it and a computer reads it Motivation: help people with severe neuromuscular disability, the ultimate means of interaction with any technology without verbal speech or typing. It's “mind reading”!
Related BCI Research ● Reading thoughts with EEG – Abject Failure, Fujimaki 1990 ● Duke ● Graz Brain-Computer Interface: uses
Modern Brain-Computer Interfaces
Neuromuscular Techniques to Read Language from the Brain ● ● The twist: use neuromuscular readings of the motor- regions associated with speech in lieu of reading the thoughts themselves Obtain this information in real-time through use of brain imaging Use a pseudo-language Graz-style if necessary – project still has other “twists” which make it worthwhile Use a neuroelectric implant if sufficient resolution cannot be obtained with brain-imaging
Language of Choice ● ● Since muscular motions of the lips will require extremely high resolution to distinguish, use a different means of communication Sign language has broad, distinct muscle activity Limit our “language” to a small set of words in a particular type of language (ALS, TLS, ULS) Identical neurolinguistic activity as verbal language ● Proof-of-concept for more refined study ● Or, perhaps it'd be useful in itself.
Interpretation of Brain Activity ● ● ● With either brain imaging or direct neuronal activity, a calibration period will be needed to map neural motor activity to particular signs The Duke monkey project mapped for 12 and 24 months per monkey. Using the robust algorithms they generated, we'll map for much less. Motor mapping software will have to be adapted from existing neuroanalysis libraries Sign language output can be translated directly using existing algorithms and software into words (Su et al. ) The final output can be converted to audio or can
Follow-up Studies ● ● The most natural is to make a “lipreader” that translates muscle motion of the lips into thoughts Such a product would be directly implementable as a commercial or medical product. Other neurotechnology studies: brain “input” devices As technology grows more refined, skip the motor cortex altogether and access the premotor cortex.
- Slides: 7