Drowsiness Detection Using Brain Signal Recognition Deep Neural Network (BSRDNN)

Abstract

A novel deep learning architecture, brain signal recognition deep neural network (BSRDNN), based on a one-dimensional convolutional neural network model (1D-CNN) and artificial neural network (ANN), is proposed for drowsiness detection from single-channel electroencephalographic (EEG) data. The effectiveness of the method is shown using the MIT/BIH polysomnographic EEG dataset (MIT/BIH-PED) with more than 80h long-term EEG data collected by a single electrode. EEG signals for 16 subjects were classified by BSRDNN as wakefulness, drowsiness, and sleep. BSRDNN was used via two approaches: Option 1 consists of feature extraction and classification by deep learning; in Option 2, feature and classification are performed by machine learning algorithms, naïve Bayes (NB), k-nearest neighbours (KNN), random forest (RF), and stochastic gradient descent (SGD). Combined-subject validation was applied extraction to enhance the performance of the proposed technique. Simulations demonstrated better performance in terms of accuracy, recall, F1-score and precision compared to the current state-of-the-art techniques applied to the same dataset: We obtained 92.31% overall accuracy in Option 1, and 94.8-100% in Option 2. The proposed novel BSRDNN model demonstrates clear superiority over those featured in published research that used the same MIT/BIH-PED dataset. It can perform its designated task with less trainable parameters and arithmetic operations compared to other models, resulting in faster training and testing phases. This enhanced speed facilitates quicker drowsiness detection, thereby reducing the overall time required for the process. ©2024 The authors.