EEG-Based Motor Imagery Classification Using Multilayer Perceptron Neural Network

Signals derived from brain activity can be used as commands to control an external device or application in Brain-Computer Interface (BCI) systems. Electroencephalography (EEG) is widely used to record brain signals due to its non-invasive nature, relatively low-cost, and high temporal resolution. BCI performance depends on choices regarding available options for signal pre-processing, classifiers, and feature extraction techniques. In this paper, we describe the use of an Artificial Neural Network (ANN) based on a Multilayer Perceptron (MLP) architecture as a classifier to identify motor imagery tasks using EEG signals from nine subjects of an experimental data set. BCIs based on brain signals recorded during motor imagery tasks use the changes in amplitude of specific cortical bands as features. Moreover, we evaluated the effect of systematically decreasing the number of inputs (EEG channels) on the classifier performance. The results show that a MLP classifier was able to segregate the EEG signature of four motor imagery tasks with at least 70% accuracy using at least 12 EEG channels.