HGRBD - A Multimodal Dataset for Human Hand Gesture Recognition and Prosthetic Control

This dataset contains synchronised multimodal recordings for hand gesture recognition and prosthetic control research. Data were collected from 27 right-handed South Asian adults performing 17 hand and wrist gestures across five recording sessions. The dataset includes three-channel forearm surface electromyography, five-finger flex sensor signals, inertial measurement unit-based hand kinematics, four force-sensitive resistor channels for forearm force myography, and galvanic skin response. The dataset is organised into participant folders named P01 through P27. Each session contains three files: an annotation CSV file, a synchronised sensor data CSV file, and an sEMG/GSR HDF5 file. The annotation file contains time_elapsed, gesture_label, and repetition, sampled at approximately 23.5 Hz. The synchronised sensor data file contains IMU acceleration columns acc_x, acc_y, and acc_z, Euler angle columns yaw, pitch, and roll, FSR columns fsr_1 through fsr_4, and finger flexion angle columns angle_thumb, angle_index, angle_middle, angle_ring, and angle_little, also sampled at approximately 23.5 Hz. The HDF5 file contains time_elapsed, emg_br, emg_ed, emg_fcr, and gsr, sampled at 5 kHz. A participant metadata file provides demographic and anthropometric information for all participants, including participant identifier, gender, age, height, weight, forearm length, forearm circumference, and wrist circumference. The dataset supports studies of multimodal sensor fusion, cross-user generalisation, session repeatability, and prosthetic hand control.