Microsleep Predicting Comparison Between LSTM and ANN Based on the Analysis of Time Series EEG Signal

Abstract

A microsleep is an unintentional, transient loss of consciousness correlated with sleep that lasts up to fifteen seconds. Electroencephalogram (EEG), recordings have been extensively utilized to diagnose and study various neurological disorders. This study analyzes time series EEG signals to predict microsleep employing two deep learning models: Long-Short Term Memory (LSTM) and Artificial Neural Network (ANN). The findings show that the ANN model achieves outstanding metrics in microsleep prediction, outperforming the LSTM in key performance metrics. The model demonstrated exceptional performance, as demonstrated by the outcomes of the Scatter Plot, R2 Score, Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE). Between the two models, the ANN model achieved the most significant R2, MAE, MSE, and RMSE values (0.84, 1.10, 1.90, and 1.38) compared to the LSTM model. The critical contribution of this study lies in its development of comprehensive and effective methods for accurately predicting microsleep events from EEG signals.