Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Modified TSception for Analyzing Driver Drowsiness and Mental Workload from EEG

Gourav Siddhad, Anurag Singh, Rajkumar Saini, Partha Pratim Roy

TL;DR

The paper tackles driver drowsiness detection and mental workload assessment using EEG by introducing a Modified TSception architecture with hierarchical temporal refinement, Adaptive Average Pooling, and a two-stage fusion to improve reliability and cross-task generalization. It demonstrates competitive SEED-VIG performance (83.46% accuracy, CI 0.24) and state-of-the-art STEW results (2-class 95.93%, 3-class 95.35%), indicating enhanced stability and cross-task applicability. The approach directly addresses inter-subject variability and real-world deployment concerns by multi-scale temporal feature extraction and adaptive pooling. These findings suggest a robust EEG-based framework capable of reliable cognitive state monitoring in safety-critical settings and across different EEG tasks.

Abstract

Driver drowsiness remains a primary cause of traffic accidents, necessitating the development of real-time, reliable detection systems to ensure road safety. This study presents a Modified TSception architecture designed for the robust assessment of driver fatigue using Electroencephalography (EEG). The model introduces a novel hierarchical architecture that surpasses the original TSception by implementing a five-layer temporal refinement strategy to capture multi-scale brain dynamics. A key innovation is the use of Adaptive Average Pooling, which provides the structural flexibility to handle varying EEG input dimensions, and a two - stage fusion mechanism that optimizes the integration of spatiotemporal features for improved stability. When evaluated on the SEED-VIG dataset and compared against established methods - including SVM, Transformer, EEGNet, ConvNeXt, LMDA-Net, and the original TSception - the Modified TSception achieves a comparable accuracy of 83.46% (vs. 83.15% for the original). Critically, the proposed model exhibits a substantially reduced confidence interval (0.24 vs. 0.36), signifying a marked improvement in performance stability. Furthermore, the architecture's generalizability is validated on the STEW mental workload dataset, where it achieves state-of-the-art results with 95.93% and 95.35% accuracy for 2-class and 3-class classification, respectively. These improvements in consistency and cross-task generalizability underscore the effectiveness of the proposed modifications for reliable EEG-based monitoring of drowsiness and mental workload.

Modified TSception for Analyzing Driver Drowsiness and Mental Workload from EEG

TL;DR

The paper tackles driver drowsiness detection and mental workload assessment using EEG by introducing a Modified TSception architecture with hierarchical temporal refinement, Adaptive Average Pooling, and a two-stage fusion to improve reliability and cross-task generalization. It demonstrates competitive SEED-VIG performance (83.46% accuracy, CI 0.24) and state-of-the-art STEW results (2-class 95.93%, 3-class 95.35%), indicating enhanced stability and cross-task applicability. The approach directly addresses inter-subject variability and real-world deployment concerns by multi-scale temporal feature extraction and adaptive pooling. These findings suggest a robust EEG-based framework capable of reliable cognitive state monitoring in safety-critical settings and across different EEG tasks.

Abstract

Driver drowsiness remains a primary cause of traffic accidents, necessitating the development of real-time, reliable detection systems to ensure road safety. This study presents a Modified TSception architecture designed for the robust assessment of driver fatigue using Electroencephalography (EEG). The model introduces a novel hierarchical architecture that surpasses the original TSception by implementing a five-layer temporal refinement strategy to capture multi-scale brain dynamics. A key innovation is the use of Adaptive Average Pooling, which provides the structural flexibility to handle varying EEG input dimensions, and a two - stage fusion mechanism that optimizes the integration of spatiotemporal features for improved stability. When evaluated on the SEED-VIG dataset and compared against established methods - including SVM, Transformer, EEGNet, ConvNeXt, LMDA-Net, and the original TSception - the Modified TSception achieves a comparable accuracy of 83.46% (vs. 83.15% for the original). Critically, the proposed model exhibits a substantially reduced confidence interval (0.24 vs. 0.36), signifying a marked improvement in performance stability. Furthermore, the architecture's generalizability is validated on the STEW mental workload dataset, where it achieves state-of-the-art results with 95.93% and 95.35% accuracy for 2-class and 3-class classification, respectively. These improvements in consistency and cross-task generalizability underscore the effectiveness of the proposed modifications for reliable EEG-based monitoring of drowsiness and mental workload.
Paper Structure (15 sections, 5 equations, 2 figures, 1 table)

This paper contains 15 sections, 5 equations, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background
  3. Methodology
  4. TSception Preliminaries
  5. Modified TSception
  6. Layer Architecture of Modified TSception
  7. Comparison of Architectural Enhancements
  8. Results and Discussion
  9. Experimental Data
  10. Classifiers
  11. Performance Evaluation
  12. Driver Drowsiness Detection
  13. Mental Workload Assessment
  14. Overall Performance and Discussion
  15. Conclusion

Figures (2)

  • Figure 1: Architectural details of the proposed Modified TSception model: (a) Overall architecture, (b) Temporal Block (Tception layer), and (c) Spatial Block (Sception layer). Dashed boxes represent the specific modifications made to the original TSception framework. Abbreviations: BN (Batch Normalization), FC (Fully Connected), Conv (Convolution), AP (Average Pooling), ADP (Adaptive Average Pooling), GAP (Global Average Pooling)
  • Figure 2: Barchart showing the performance comparison of different methods on the SEED-VIG for driver drowsiness detection and STEW dataset for mental workload assessment using 5-fold cross-validation.