Inertial Dataset for Posture Recognition in Agriculture and Construction Tasks

This dataset, manually labeled, contains 10 hours and 40 minutes (60 Hz) of 8 typical working posture classes (standing, reaching, stooping, squatting, kneeling, lifting/lowering, carrying, and others), acquired with 16 subjects in three distinct scenarios in a lab environment:

Isolated postures or short sequences without any associated task;

Agriculture task (bricklaying) circuit;

Construction task (harvesting) circuit.

Two full-body inertial motion caption systems (17 Xsens MTw Awinda IMUs each, from Xsens Technologies, B.V., The Netherlands) were used, connected to, respectively:

Xsens MT Manager, providing raw inertial data (acceleration, angular velocity, and magnetic field data - csv files);

Xsens MVN Analyze, providing processed data (quaternions, Euler angles, position, linear velocity, acceleration, angular velocity, angular acceleration, joint angles, ergonomic angles, center of mass, and magnetic field - xlsx files).

More information about the dataset acquisition and organization is detailed in readme.pdf file. For any questions, please contact Diogo R. Martins at [e-mail] or Sara M. Cerqueira at [e-mail].