Selected work

Representative Papers

Pupillometry and Brain Dynamics for Cognitive Load in Working Memory

Feb 11, 2026

This study addresses the need for accurate, lightweight, and wearable assessment of working memory cognitive load to support adaptive learning, clinical monitoring, and brain–computer interface applications. Leveraging the OpenNeuro “digit span task” dataset, the authors integrate pupillometry and electroencephalography (EEG) signals within a physiologically interpretable, lightweight classification framework. This framework employs Catch-22 time-series feature extraction, conventional machine learning models, and SHAP-based interpretability analysis. Results demonstrate that feature-based approaches outperform deep learning in both binary and multiclass cognitive load classification tasks. Notably, pupillometry alone achieves performance comparable to EEG, enabling efficient and portable cognitive load recognition. These findings establish a novel paradigm for wearable cognitive monitoring that balances accuracy, interpretability, and practicality.

0 citationsRead paper

Cross-Modal Computational Model of Brain-Heart Interactions via HRV and EEG Feature

Jan 11, 2026arXiv.org

This study investigates whether electrocardiogram (ECG) signals can serve as a reliable surrogate for electroencephalogram (EEG) in monitoring cognitive load outside controlled laboratory settings. To this end, the authors propose a cross-modal regression framework that maps ECG-derived features—specifically heart rate variability (HRV) and Catch22 time-series descriptors—to EEG band power metrics using XGBoost. To address the scarcity of real-world HRV data, they introduce PSV-SDG, a novel method for generating synthetic HRV sequences, which is shown to significantly enhance model performance. The resulting approach enables lightweight, interpretable, and robust modeling of cognitive states using only wearable ECG devices. Experimental results demonstrate that ECG-based features effectively proxy EEG-derived cognitive indicators, thereby establishing a foundation for low-cost, real-time cognitive monitoring in elderly and clinical populations.

0 citationsRead paper

Unveiling the Heart-Brain Connection: An Analysis of ECG in Cognitive Performance

Jan 04, 2026arXiv.org

This study investigates whether electrocardiogram (ECG) signals can serve as a viable alternative to electroencephalography (EEG) for real-time, wearable cognitive load monitoring. By simultaneously acquiring multimodal physiological data during working memory and passive auditory tasks, the authors extract time-domain heart rate variability (HRV) and Catch22 time-series features from ECG signals and develop a cross-modal XGBoost mapping framework to project these ECG-derived features into the EEG-based cognitive representational space. This work presents the first systematic validation of ECG’s proxy capability in cognitive load assessment, demonstrating that ECG-derived features effectively capture dynamic changes in cognitive states and achieve strong performance in classification tasks. The findings establish a novel, interpretable, real-time, and practical paradigm for physiological computing in everyday wearable applications.

0 citationsRead paper

Recent publications

Latest Papers

Pupillometry and Brain Dynamics for Cognitive Load in Working Memory

Feb 11, 2026

This study addresses the need for accurate, lightweight, and wearable assessment of working memory cognitive load to support adaptive learning, clinical monitoring, and brain–computer interface applications. Leveraging the OpenNeuro “digit span task” dataset, the authors integrate pupillometry and electroencephalography (EEG) signals within a physiologically interpretable, lightweight classification framework. This framework employs Catch-22 time-series feature extraction, conventional machine learning models, and SHAP-based interpretability analysis. Results demonstrate that feature-based approaches outperform deep learning in both binary and multiclass cognitive load classification tasks. Notably, pupillometry alone achieves performance comparable to EEG, enabling efficient and portable cognitive load recognition. These findings establish a novel paradigm for wearable cognitive monitoring that balances accuracy, interpretability, and practicality.

0 citationsRead paper

Cross-Modal Computational Model of Brain-Heart Interactions via HRV and EEG Feature

Jan 11, 2026arXiv.org

This study investigates whether electrocardiogram (ECG) signals can serve as a reliable surrogate for electroencephalogram (EEG) in monitoring cognitive load outside controlled laboratory settings. To this end, the authors propose a cross-modal regression framework that maps ECG-derived features—specifically heart rate variability (HRV) and Catch22 time-series descriptors—to EEG band power metrics using XGBoost. To address the scarcity of real-world HRV data, they introduce PSV-SDG, a novel method for generating synthetic HRV sequences, which is shown to significantly enhance model performance. The resulting approach enables lightweight, interpretable, and robust modeling of cognitive states using only wearable ECG devices. Experimental results demonstrate that ECG-based features effectively proxy EEG-derived cognitive indicators, thereby establishing a foundation for low-cost, real-time cognitive monitoring in elderly and clinical populations.

0 citationsRead paper

Unveiling the Heart-Brain Connection: An Analysis of ECG in Cognitive Performance

Jan 04, 2026arXiv.org

This study investigates whether electrocardiogram (ECG) signals can serve as a viable alternative to electroencephalography (EEG) for real-time, wearable cognitive load monitoring. By simultaneously acquiring multimodal physiological data during working memory and passive auditory tasks, the authors extract time-domain heart rate variability (HRV) and Catch22 time-series features from ECG signals and develop a cross-modal XGBoost mapping framework to project these ECG-derived features into the EEG-based cognitive representational space. This work presents the first systematic validation of ECG’s proxy capability in cognitive load assessment, demonstrating that ECG-derived features effectively capture dynamic changes in cognitive states and achieve strong performance in classification tasks. The findings establish a novel, interpretable, real-time, and practical paradigm for physiological computing in everyday wearable applications.

0 citationsRead paper