Exercise-specific joint selection shows limited gains in classification
When Does Exercise-Specific Joint Selection Help? An Audit of Evaluation and Control Design
Artificial Intelligence
Summary
This paper looks at whether choosing specific body joints for recognizing exercise movements improves accuracy. The authors studied over a thousand exercise repetitions and found that picking particular joints gives only small or no clear benefits depending on how the tests were done. They also tested different ways to compare results and found some approaches work better than others, but no new methods or clinical advantages were established. Their work highlights the need for careful testing methods when claiming that joint selection helps.
What this means in practice
- •For fitness app developers: Improve movement classification algorithms by applying stringent evaluation controls for joint selection to avoid overestimating benefits.
- •For physical therapy equipment makers: Use robust testing methods to assess sensor placement and joint selection for devices that monitor exercise correctness.
Tested on one dataset.
Authors
Haotian Chen, Jingkun Yu, Yuning Zhang, Bowen Ye
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.