Neural networks for robust classification of mental tasks

This paper investigates appropriate neural classifiers for the recognition of mental tasks from on-line spontaneous EEG signals. The classifiers are to be embedded in a portable brain-computer interface called ABI. We evaluate different kinds of classifiers, from statistical approaches to neural networks, with 8 healthy persons. Subjects' performance is analyzed off-line and, for three of them, also on-line in the presence of biofeedback. The proposed ABI robustly recognizes three mental tasks from on-line spontaneous EEG signals. Correct recognition is around 70%. This modest rate is largely compensated by two properties of ABI: wrong responses are below 5% and it makes decisions every 1/2 second. Also, since the subject and his/her personal ABI learn simultaneously from each other, subjects master it rapidly: one of the subjects achieved excellent control in just 5 days of training. Analysis of learned EEG patterns confirms that for a subject to operate satisfactorily an ABI, the latter must fit the individual features of the former. Building individual interfaces greatly increases the likelihood of success, as demonstrated for all subjects we have worked with despite the short training time of most of them.