EEG-Driven Physical Load Detection During Human Walking for Robotic and Automation Systems

¹ Department of Robotics & Artificial Intelligence, School of Electrical and Electronics Engineering, SASTRA Deemed to be University, Thanjavur, Tamil Nadu, India
² School of Electronics Engineering, Vellore Institute of Technology, Chennai, Tamil Nadu, India
³ Institute of Research and Development, Duy Tan University, Da Nang, Vietnam
⁴ School of Engineering & Technology, Duy Tan University, Da Nang, Vietnam
⁵ Biomedical Engineering Research Group, School of Engineering, University of Leicester, Leicester LE1 7RH, United Kingdom
⁶ School of Electrical and Electronics Engineering, SASTRA Deemed to be University, Thanjavur, Tamil Nadu, India
⁷ Chief Technical Officer, Qneuro India Private Limited, Chennai, Tamil Nadu, India

^* Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Published online: 29 April 2026

Abstract

Understanding neural responses to varying physical loads is essential for developing ergonomic designs. Conventional methods for analyzing peripheral muscle activity, such as electromyography (EMG) and kinematic analysis, provide only limited insight into the cortical dynamics associated with physical tasks. To overcome this limitation, the present study introduces an electroencephalography (EEG)-based approach to investigate brain activity during load-bearing conditions. Participants performed a 100-meter walking task while carrying a 5 kg shoulder load, during which raw EEG signals were recorded. These signals were transformed using Continuous Wavelet Transform (CWT) to generate scalograms, capturing both temporal and frequency-domain characteristics of neural activity. Deep learning (DL) models were then trained, validated, and tested using these representations, and their performance was evaluated through standard metrics. Several DL architectures, including CNN, ResNet18, VGG19, DenseNet, and ResNet50, were employed to extract spatial–temporal features associated with load conditions. Among these, ResNet18 achieved the highest accuracy of 66.83%, outperforming conventional feature-based approaches. Additionally, the occipital cortex showed the highest classification accuracy (69.09%) in distinguishing between no-load and 5 kg load conditions. These findings highlight the potential of DL-based EEG analysis for workload monitoring, fatigue assessment, and brain–computer interface applications.