What question did this study set out to answer?

This study evaluates self-supervised learning (SSL) for enhancing gait event detection using inertial sensors.

November 11, 2025Applied SciencesOpen Access

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

View Full Paper

Ask AI

Bookmark

Authors

In silico analysis shows self-supervised learning improves accuracy and efficiency in gait detection using inertial sensors for mobile health.

This study evaluates self-supervised learning (SSL) for enhancing gait event detection using inertial sensors.
Analysis of six public IMU datasets for gait detection
Contrastive self-supervised learning employed with sensor-aware augmentations
Models pre-trained on unlabeled data, fine-tuned with limited labeled data
Systematic assessment of label efficiency and performance under device shifts
Comparison of model architectures including CNN, TCN, BiLSTM, and Transformer.
SSL outperformed supervised methods in gait detection accuracy
Pretraining improved accuracy by ΔF1 +0.08–0.15 with reduced stride-time error by −8 to −12 ms
SSL reduced required labeled data by up to 95%
Enhancements drove by sensor-aware augmentations
Robust performance across varying sampling rates and device placements.

Mănescu et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark

In silico analysis shows self-supervised learning improves accuracy and efficiency in gait detection using inertial sensors for mobile health.