Datasets for Bilateral Lower-Limb Neuromechanical Signals in Able-Bodied and Impaired Individuals with Wearable and Ambient Sensors (BLISS)

Data collection method:

The subject's weight and height were measured using analog devices. The motion capture cameras and force plates were calibrated prior to the experiment. 15 wearable sensors (inertial measurement (IMU) plus electromyography (EMG) integrated units) were attached to the participants' lower limbs. 26 passive reflective markers were attached non-invasively on top of skeletal landmarks according to the IOR lower body automated identification of markers (AIM) model. The attachment of sensors and markers is visually illustrated in the additional metadata files associated with the dataset. The walking procedure involved moving forward for about 5 meters. Each subject performed this trial approximately 50 times, with administrated breaks to prevent fatigue. The number of gait bouts in each direction was roughly equal, with each bout covering nearly 5 meters. Subjects walked at their self-selected speeds, and trials with pauses or trips within the force plate area were discarded. Data collection took about 3 hours for healthy subjects and extended to 4-5 hours for those with severe impairments.

Data processing and preparation activities:

The dataset was annotated using a visual3D. EMG and IMU signals were filtered thoroughly according to literature recommendations. Visual3D software analyzed raw Qualysis signals, assigning joint reference frames and calculating kinematic and dynamic data. The processed data was later segmented into analysis windows to create additional feature-based datasets. Each window's features were computed, including 780 features from 30 sensors. MATLAB was used for gait analysis, comparing joint kinematic/kinetic data to normative data and processing EMG envelopes and gait cycle data. Data from the test subjects was completely anonymized using unique codes.

Technical details and requirements:

Qualysis software (Qualysis, Göteborg, Sweden) 2023; Visual3D software (HAS-motion; Kingston, Ontario) 2023; MATLAB software (Mathworks, Natick, Massachusetts, United States) 2022; 15 Trigno Avanti™ Sensors; Delsys, Natick, MA, USA. The motion capture system used (Qualysis, Göteborg, Sweden) includes 14 cameras (2 Miqus and 12 Arqus cameras) to track passive markers. Ground reaction forces are recorded using four force plates (BMS400600; Amti, Watertown, MA, USA) fixed to the lab floor.