• Kratika Saxena
• Indu Shrivastava
• Nitya Khare
• Swati Khanve

Fatigue is a critical factor affecting cognitive performance, decision-making, and overall well-being, particularly in high-risk domains such as transportation, healthcare, and industrial operations. Traditional fatigue detection methods rely on subjective self-assessment or physical indicators, which may not accurately reflect mental fatigue levels. In this study, we propose a Convolutional Recurrent Neural Network (CRNN)-based model for fatigue mental state prediction using Electroencephalogram (EEG) signals. The model leverages Convolutional Neural Networks (CNNs) for spatial feature extraction and Recurrent Neural Networks (RNNs) to capture temporal dependencies in EEG time-series data. The proposed LogMel-CRNN model integrates Short-Time Fourier Transform (STFT) and Mel spectrogram transformations to convert EEG signals into spectrogram representations, enhancing feature extraction. The 1D-CNN layers extract spatial patterns from EEG spectrograms, while the Long Short-Term Memory (LSTM) or Gated Recurrent Unit (GRU) layers process temporal dependencies, allowing the model to capture fatigue-related transitions over time. The final classification layer utilizes a softmax activation function for multi-class fatigue classification or a sigmoid function for binary classification (fatigue vs. alert state).The experimental evaluation, conducted on publicly available EEG fatigue datasets, demonstrates the superior performance of the proposed model compared to traditional machine learning methods. Results show that the LogMel-CRNN model achieves an accuracy improvement of 15–20% over conventional EEG-based classifiers. Additionally, the integration of attention mechanisms further enhances the model’s ability to focus on critical temporal segments related to fatigue onset. The study also investigates the impact of filter size variations in the convolutional layers, demonstrating improved feature extraction with optimized configurations. The proposed CRNN-based approach provides a scalable and real-time fatigue detection system, making it suitable for deployment in driver monitoring systems, workplace safety applications, and cognitive workload assessment tools. Future research directions include optimizing the model for real-time embedded systems, enhancing generalization across diverse EEG datasets, and incorporating domain adaptation techniques to improve cross-subject fatigue detection accuracy.

Fatigue Detection, Mental State Prediction, Convolutional Recurrent Neural Network (CRNN), Electroencephalogram (EEG), Short-Time Fourier Transform (STFT), Mel Spectrogram, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Attention Mechanism, Cognitive Workload Assessment, Real-Time Monitoring.

"Fatigue Mental State Prediction using Convolutional Recurrent Neural Network ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 5, page no.d485-d489, May-2025, Available :http://www.jetir.org/papers/JETIR2505396.pdf

"Fatigue Mental State Prediction using Convolutional Recurrent Neural Network ", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 5, page no. ppd485-d489, May-2025, Available at : http://www.jetir.org/papers/JETIR2505396.pdf