EEG Signals for Robust Cognitive Workload Recognition Under Noisy Conditions.zip

<b>EEG Dataset for Realistic Workload Analysis</b>This dataset contains EEG recordings from 40 participants engaged in puzzle-solving and arithmetic tasks, designed to simulate real-world cognitive workload scenarios. Unlike traditional lab-controlled datasets, this collection deliberately includes a broad range of physiological and environmental artifacts, providing a realistic testing ground for robust EEG-based algorithm development.Key Characteristics:<b>Noise Sources Captured:</b> ECG (cardiac activity), EMG (muscle movements), EOG (eye blinks/movements), respiratory artifacts, and power line interference.<b>Hardware Used:</b> EEG signals were acquired using the KT88-3200 system with 16 channels at a sampling rate of 200 Hz.<b>Participant Background:</b> All participants had prior experience with dual-stage mental arithmetic protocols.Task and Recording Design:<b>Resting Phase:</b> Approximately 5 minutes, with ~1-minute segments extracted from early and late intervals.<b>Working Phase:</b> Approximately 10 minutes, also segmented into ~1-minute epochs from early and late stages.<b>Transition Break:</b> A 5-minute interval was introduced between phases to simulate workload transitions.<b>Total Duration:</b> Each participant contributed approximately 20 minutes of EEG data.<b>Session Randomization:</b> Session order was randomized to mitigate sequencing effects.File Naming Convention:Each EEG recording is stored in EDF format using the structure:<code>"SubjectXX_Y.edf"</code><code>XX</code> – Participant number (e.g., 00 to 39)<code>Y</code> – Task condition:<code>0</code>: No Workload (Resting State)<code>1</code>: WorkloadThis naming scheme ensures easy identification of each subject’s session and workload state.Participant Grouping:An accompanying file named <code><strong>Ratings (Good or Bad).txt</strong></code> is provided to indicate subjective performance ratings:<b>Good Counters:</b> 18 participants identified as having strong task performance<b>Bad Counters:</b> 22 participants with relatively lower performanceThese classifications can be used for subgroup analysis, model generalization studies, or performance-based stratification.Applications:This dataset supports research in:EEG signal denoising and artifact rejectionCognitive workload classification and regression modelingBrain–computer interface (BCI) validation under field conditionsReal-world mental workload monitoring and assessmentBy combining structured cognitive protocols with realistic environmental noise, this dataset offers a practical benchmark for evaluating EEG-based technologies in naturalistic settings.