Motor imagery EEG signal classification using long short-term memory deep network and neighbourhood component analysis

Brain Computer Interface is a technique used to measure brain activity in terms of electrical signals. The recorded Electroencephalograph (EEG) signal is highly sensitive to human activity. Therefore EEG pre-processing is necessary before important information can be extracted for EEG signal, after pre-processing important features can be easily extracted. For the purpose of minimizing the features data, Principal component analysis could be a better option. Finally, EEG classification is completed with the application of deep neural network while considering both mean square and cross entropy. The use of wrapper method for important features selections is vital as it reduces processing time and also avoid over fitting of deep neural networks (DNN), as over fitting of DNN leads to decrease in accuracy. In this work alpha and beta waves are considered. To view the usefulness of long short term memory (LSTM) DNN, for pre-processing and feature extractions of EEG basic methods are considered. LSTM DNN accuracy of more than 95 percent is achieved.