I would like to present the work done regarding how one-class recognition of cognitive brain functions across multiple subjects can be performed at the 90% level of accuracy via an appropriate choices of features which can be chosen automatically. This work extends one class work done by Larry Manevitz (Haifa University) and David Hardoon (UCL) who showed that such classification was first shown to be possible in principle albeit with an accuracy of about 60%. It also is comparable to work of various groups around the world e.g. Michell et al which have concentrated on two-class classification. The importance of this work is that one class is often the appropriate classification setting for identifying cognitive brain functions. The methodologies for one class classification used in this paper are the compression neural network for one class classification, a wrapper approach and the genetic algorithm for feature selection. In addition, versions of one-class SVM due to were investigated.