Investigating Time Domain Features in Activity Recognition for Upper Limb Rehabilitation Using Skeleton Tracking
Bibliographic record
Abstract
Accurate recognition of activities is crucial for effective tele-rehabilitation, leading to improved patient outcomes. In this study, we investigated different feature selection techniques for upper limb activity recognition. We asked 14 healthy participants (age: 24.71±2.34 years, height: 1.71±0.08 m, and BMI: 23.55±2.85 kg/m2) to perform 7 upper limb exercises for 3-5 times. During the exercises, RGB data was recorded. We extracted 2D coordinates (x,y) of 6 upper body joints as well as elbow and shoulder joint angle signals using AlphaPose technique. To improve data quality, we pre-processed the signals using the Z-score method, median filter, and moving average filter. To find significant features, we used three different feature selection techniques: 1) Recursive Feature Elimination with Random Forest, 2) Least Absolute Shrinkage and Selection Operator (LASSO), and 3) Feature Importance from Random Forests. Our results indicated that AlphaPose achieved an average joint detection and positioning score of 84% which demonstrates its effectiveness in this application. Notably, features extracted from the x-coordinate of wrist and elbow angle emerged as significantly influential in recognizing the type of exercise performed. We conducted artificiall intelligence models and in 10-fold cross-validation, STGCN++ achieved the highest accuracy (94.35 ± 3.14%) and F1 (94.30 ± 3.15%), followed by XGBoost and Random Forest. All models had high specificity (≥97.8%) and strong sensitivity (>94%). LOSO cross-validation confirmed STGCN++ as the most robust (accuracy: 87.95%, F1: 85.90%), with all models maintaining high specificity but variable sensitivity, highlighting STGCN++’s generalizability to unseen subjects. Overall, although deep learning techniques performed best, careful feature engineering can achieve comparable results with lower complexity.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".