Problem
Research questions and friction points this paper is trying to address.
joint selection
skeleton-based classification
evaluation audit
rehabilitation exercise
When Does Exercise-Specific Joint Selection Help? An Audit of Evaluation and Control Design
🤖 AI Summary
This study addresses the unclear practical gains and statistical significance of motion-specific joint selection in skeleton-based action classification. Leveraging the REHAB24 dataset, the authors systematically decouple and evaluate feature aggregation, subset structures, and temporal representations through structured controlled experiments employing kNN, RBF-SVM, logistic regression, and random convolutional networks with cross-validation. The findings reveal that random mappings can match the effectiveness of manual feature selection, underscoring the importance of explicit estimators. Notably, certain performance gains are statistically insignificant or even negative, indicating that purported clinical benefits require rigorous validation. As no new algorithmic superiority is established, this work provides a critical methodological caution for the field regarding the evaluation of skeletal representation strategies.
📝 Abstract
Exercise-specific joint selection can improve skeleton-based correctness classification, but what does that gain establish? We audit 1,057 repetitions from ten REHAB24-6 subjects, separating evaluation aggregation, subset structure, and temporal representation. The manual-subset kNN gain changes from 0.055 for pooled out-of-fold AUROC to 0.020 for equal-weight within-person AUROC; both paired intervals include zero. Among 1,000 dimension-matched random maps, 14 match or exceed the manual pooled result, versus 145 when bilateral structure and trunk inclusion are also matched. RBF-SVM retains a positive within-person gain, whereas logistic regression and a random-convolution comparator have negative point gains under that estimand. Sequence-order and paired-seed controls further qualify the interpretation. This exploratory audit shows why joint-selection claims require explicit estimands and structurally appropriate controls; it does not establish a new algorithm or clinical benefit.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Joint Selection
Evaluation Audit
Estimand
Random Mapping Controls
Skeleton-based Classification
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
No related jobs found.
