Building dependable EEG classifiers for the real world - It's not just about the hardware

Abstract

One of the major deficiencies with the EEG-based classifiers used in today's laboratory settings is that they are often ill suited for the real world. In many cases the classifiers that were painstakingly developed in the controlled laboratory environment become unreliable with increased mobility of the user. In addition to increased mobility, many real world scenarios impose constraints on data collection that cannot be accommodated by the lab-created classifier. Addressing these issues throughout the development process of EEG-based classifiers by building hardware, software, and algorithms intended for use in the real world should result in more dependable classifiers. With this approach we were able to collect and classify data on a research vessel at sea, in the desert by night, on dismounted soldiers in the training field, and everywhere between. © 2009 Springer.