Real-Time Cognitive Load Estimation via Wearable EEG - Interactive Simulation and Classification Prototype

We present a fully interactive prototype for real-time cognitive load estimation using wearable EEG. The system streams multichannel EEG data, applies on-the-fly preprocessing (1\u201345 Hz bandpass, 50 Hz notch), and buffers sliding windows (1\u20133 s) for classification by a lightweight 1D-CNN. A responsive Streamlit GUI allows users to adjust window length, step size, and injected noise; visualize raw signals and confidence history; monitor inference latency (<50 ms average); and control playback across multiple EEG files. The prototype achieves near-perfect test accuracy on a synthetic cognitive-load dataset, demonstrates robustness to added noise, and provides detailed latency and confidence metrics in real time. We release a modular training script (train_cnn_eeg_classifier.py) and an interactive demo (app.py), along with full documentation and example recordings. This work lays the foundation for wearable, adaptive human\u2013machine interfaces that can sense and respond to cognitive workload in realistic settings.