Energy-Time Profiling for Machine Learning Methods to EEG Classification

The Electroencephalography discipline studies a type of signals called Electroencephalograms (EEGs), which represent the electrical activity of different parts of the brain. EEGs are composed of a massive number of features that could be used to create an intelligent recognition system. Nevertheless, the high number of available features difficult the correct classification of the signals as most of them do not contain relevant information. The use of Feature Selection (FS) techniques allows the reduction of the number of features by extracting a reduced but powerful feature subset. Afterwards, the problem can be addressed by using both the reduced features along with classification methods. In this sense, this work proposes a comparison of five supervised classifiers to solve an Electroencephalography problem in the context of Motor Imagery-based BCI tasks. The predictive models are evaluated in terms of classification rate and performance (execution time and energy consumption), with the idea of determining which alternative offers the best trade-off among all the objectives.