A Multimodal Biometric Dataset for Signature Verification (EEG+EMG) V1.0

This publicly available multimodal dataset contains synchronized electroencephalography (EEG) and electromyography (EMG) recordings from 30 healthy right-handed adults (25 male, 5 female, aged 22-30 years) during three signature-related tasks designed for biometric authentication and brain-computer interface research. The dataset captures neural and muscular dynamics during: (1) mental imagery of signature visualization, (2) motor imagery of signature writing, and (3) physical execution of signatures using pen and paper. Data was acquired in a controlled environment using a g.HIamp amplifier with 5 EEG channels (AF3, AF4, T7, T8, Pz at 512 Hz) and 3 EMG channels (Extensor Digitorum, Extensor Carpi Radialis Longus, Pronator Teres with reference electrode). Each participant completed 10 trials (5 genuine and 5 forged signatures) following standardized auditory instructions, with task phases marked by manual annotations. Raw data is provided in MATLAB (.mat) format containing 5×N (EEG) and 4×N (EMG) matrices, while preprocessed versions include downsampled (128 Hz), filtered EEG (.csv) and rectified EMG (.csv) data. The dataset features a consistent naming convention (e.g., P_eeg_15G(2).mat where P=raw, 15=subject ID, G=genuine, (2)=trial 2) and organized folder structure separating raw and preprocessed files by modality. Ethical approval was obtained from CUET, with all data anonymized and shared under CC-BY 4.0 license. This resource supports research in biometric systems, motor control studies, and multimodal machine learning applications.