October 10, 2025Open Access

Self-Supervised Gait Event Detection from Smartphone IMUs for Human Performance and Sports Medicine

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Self-supervised learning improves gait event detection accuracy in mobile health applications, suggesting a scalable approach for clinical use.

Self-supervised learning consistently outperforms supervised baselines in gait event detection across multiple datasets.
Pretraining with self-supervised learning improved accuracy by ΔF1 +0.08–0.15 while reducing stride-time error by 8–12 ms.
Utilizing sensor-aware augmentations led to the strongest transfer gains in performance, particularly through axis-swap and drift.
CNNs and TCNs provided the best trade-offs between efficiency and accuracy, demonstrating robustness across various devices and sampling rates.

Tudor et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

Self-supervised learning improves gait event detection accuracy in mobile health applications, suggesting a scalable approach for clinical use.