Enhanced Modified Z-Score-Based EEG Signal Preprocessing for Driver Fatigue Classification Using a DBN-LSTM Hybrid Deep Learning Model

Abstract

This study presents a method for driver fatigue detection using EEG signals, combining an enhanced modified Z-score-based preprocessing technique with a hybrid deep learning model that integrates Deep Belief Networks (DBN) and Long Short-Term Memory (LSTM) networks. The enhanced modified Z-score preprocessing method effectively reduces noise and outliers in EEG data, significantly improving the quality of features for fatigue classification. The DBN component is used for unsupervised feature extraction, while the LSTM component captures temporal dependencies in the data, enhancing the accuracy of the model. Experimental results showed that the proposed model achieved an overall accuracy of 82.41%, a specificity of 65.10%, and an F1-score of 84.90%, indicating robust performance in classifying different driver fatigue states. These findings demonstrate that the DBN-LSTM hybrid model, combined with the enhanced preprocessing technique, offers a promising solution for real-time driver fatigue detection, with potential applications in critical areas such as driver monitoring systems and industrial safety.