ACM TechNews

Standardizing the Brain-Machine Interface

IEEE Spectrum (04/08) Vol. 45, No. 4, P. 16; Peck, Morgen E.

Operating a neural prosthetic requires algorithms to decode brain signals in order to drive whatever device the subject is attempting to move or manipulate by thought, and some people in the field of brain-machine interface research say the time has come to develop a standard decoder algorithm. One project in this vein is an effort led by MIT computer scientist Lakshminarayan Srinivasan, whose goal is to unify elements from all algorithms designed at major brain-machine interface labs in an attempt to create a generic approach that supports and augments each design. Approaches to neural prosthetics have varied across different projects, with some interfaces featuring direct brain or skull implants and others featuring electrodes attached to the scalp. Srinivasan's algorithms have performed as well or better than those he sought to bring together, at least in simulation. But simulation is no measure for evaluating an algorithm, says Duke University Medical Center engineer Mikhail Lebedev. Brain-machine interfaces involve the adaptation of the brain and the algorithms to one another. The brain also learns, to a certain degree, how to game the algorithms' rules to achieve the desired outcome, so the algorithms' performance cannot be completely predicted. Srinivasan says he is currently learning electrophysiology methods and will soon begin human tests of his algorithms.

http://www.spectrum.ieee.org/apr08/6105

---

© Copyright 2008 Information, Inc. This service may be reproduced for internal distribution.