Teyj Krishnan

Brain-Computer Interface-based Motor Imagery EEG Classification - Linear Discriminant Analysis and Multi-Layer Perceptron

Started Mar. 20, 2023

Abstract or project description

For individuals who suffer from neurodegenerative diseases, motor inhibitors, such as paralysis can occur. These individuals can often lose the opportunity to engage in society like they should be able to. Therefore, it would be extremely pivotal if this movement could be restored. Scientists are aiming to reproduce the movement artificially, by reading the individual's EEG signals, and then translating them into the control of a prosthetic or external device. In order for this to happen, EEG signals must be accurately classified, in order to determine what movement they should lead to a recreation of. This study provides a review of two methods of EEG signal classification – Linear Discriminant Analysis (LDA) and a Multi-Layer Perceptron (MLP) Neural Network – in the context of Brain-Computer Interface (BCI) systems. Because EEG data is inherently high-dimensional and complex, the ability to accurately distinguish motor imagery classes is crucial for improving BCI applications. The results demonstrate that the MLP classifier achieves a significantly higher classification accuracy (99.63%) in comparison to the LDA classifier (64.15%). This substantial difference in classification accuracy highlights the effectiveness of neural networks in handling complex data and may have the potential to enhance future BCI performances, holding great promise for individuals, specifically those with motor-inhibiting neurodegenerative diseases, who depend on these systems for an improved quality of life.