Robust Analysis of Deep Learning Algorithm for Brain Computer Interface (BCI) Systems based on EEG Signals

Abstract

The electroencephalogram (EEG) signals have been used for a variety of neurological evaluation applications, and they are useful for patients undergoing epileptic diagnosis since they may identify deficits, illnesses, and diseases related to human brains. The neurological conditions of patients with epileptic seizures were examined in this study to get an objective and precise previous diagnosis. The electrical activities of the human mind and bodily functions associated with the nervous system are contained in the EEG signal. These EEG data were gathered from signals of a sort that were previously captured and uploaded datasets; many of these datasets are difficult to interpolate using the currently used techniques. Certain epileptic seizure behaviors and trustworthy diagnoses have suggested that the urgency of therapy be confiscated in this inquiry. A training model for classification is a recurrent convolutional neural network (RCNN), and a preprocessing with autoencoder. It is a fully supervised technique for predicting epileptic seizures and produces excellent results, such as an increase in accuracy of 98.78, specificity of 92.32, and sensitivity of 98.52. In comparison to other classifiers, the research demonstrates that the RCNN model is the best and provides great performance metrics. Our work helps epileptic patients in unusual circumstances get a better diagnosis and therapy.