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

To address the challenges in human activity recognition (HAR) and advance research in lower-limb assistive devices and machine learning (ML) for HAR, we introduce the BLISS dataset (Bilateral Lower-Limb Neuromechanical Signals). This benchmark dataset includes bilateral EMG and limb kinematics from wearable sensors for 21 subjects, encompassing both healthy individuals and those with mild to severe gait abnormalities. The selected abnormalities, which are clinically prevalent, include reduced knee flexion due to stiffness and/or overweight conditions, and dorsiflexor weakness resulting from stroke, multiple sclerosis and lumbar radiculopathy. Subjects perform free ground-level walking in an uncontrolled environment across multiple trials, each representing a complete gait bout. Additionally, marker-based motion capture and force plates provide ground truth estimates of limb positions and ground reaction forces. Comprehensive analyses estimate angular accelerations, velocities, positions, and torques at individual joints. The dataset is fully annotated for gait cycle phases: loading response (LR), mid-stance (MST), terminal stance (TS), pre-swing (PSW), and swing (SW),. This dataset complements existing benchmarks by offering detailed guidelines for sensor modality selection, analysis, and annotation, and balancing data between healthy and impaired subjects.