Interpretable Side-Aware Gait-State Representations for Post-Stroke Adaptive Neurorobotics

A secondary analysis of a public multimodal gait dataset comprising 138 able-bodied adults and 50 adults with stroke. The study derived three interpretable gait states from kinematic and sEMG waveforms for a strict complete-case stroke cohort of 43 individuals. This dataset, authored by Rocco Salvatore Calabrò and shared under a CC-BY-4.0 license, presents a hypothesis-generating framework for neurorobotic state-representation design.

Use Cases

• Designing adaptive-assistance algorithms based on side-aware multimodal gait-state representations.
• Comparing paretic and non-paretic limb biomechanics using derived side-difference views.
• Validating gait-state clustering methods using fused kinematic and sEMG waveform data.
• Investigating the contribution of specific domains like ankle angle or vastus lateralis sEMG to gait-state classification.

Strengths

• Derived from a public multimodal dataset with 188 total participants (138 able-bodied, 50 with stroke).
• Analysis is based on 11 waveform domains, each represented by 1,001 time-normalized points.
• The retained three-state solution was evaluated with metrics like silhouette score and bootstrap adjusted Rand index.
• Sensitivity analyses, such as erector-spinae exclusion, were performed to assess result stability.

Limitations

• The dataset is very small at 29.6 KB, indicating it contains processed results or a document, not the raw source data.
• Column-level documentation is absent; field semantics must be inferred after download.
• The row count for the underlying data is unknown, which may limit suitability assessment.

Provenance

Source

Secondary analysis of a public multimodal gait dataset.

Collection Method

Functional principal component analysis and clustering (Ward hierarchical, K-means) on kinematic and sEMG waveforms.

Freshness

Last updated 2026-05-25 06:05:44; freshness should be verified.