Foot-mounted IMUs provide the highest accuracy for running temporal phase estimation, with the lowest median error (0 ms) and interquartile range (5 ms for stride, 22 ms for stance), and stance estimation performance was significantly affected by running speed across 36 algorithms tested.
Key Findings
Results
Foot-mounted IMUs provided the highest accuracy for both stride and stance temporal parameter estimation compared to shank- and lower back-mounted devices.
Foot-mounted IMUs achieved a median error of 0 ms and interquartile range of 5 ms for stride estimation.
For stance estimation, foot-mounted IMUs achieved a median error of 0 ms and interquartile range of 22 ms.
Performance ranking from best to worst was: foot-mounted, shank-mounted, then lower back-mounted IMUs.
36 algorithms were compared across the three device positions.
Results
Stance estimation performance was significantly affected by running speed, whereas stride estimation was comparatively more robust to speed changes.
Three running speeds were tested: 15, 20, and 25 km/h.
The effect of speed on stance estimation was statistically significant, highlighting the importance of accounting for speed when selecting an algorithm.
The study recommends that running speed be considered as a key factor when choosing an IMU-based method for stance phase estimation.
Stride estimation showed lower interquartile range errors (5 ms) compared to stance estimation (22 ms) for the best-performing foot-mounted devices.
Methods
A total of 36 algorithms for stance and stride estimation were evaluated across three IMU positions and three running speeds in a controlled in-field study.
Ten participants performed running trials at 15, 20, and 25 km/h.
A total of 1034 strides were analyzed across all conditions.
A photocell system served as the reference system for ground truth comparison.
Algorithms varied based on device position (foot, shank, lower back), type of inertial signal analyzed, and signal processing/parameter extraction methods.
Background
A comprehensive comparison of 36 IMU-based algorithms revealed that no single universal guideline previously existed for method selection across different experimental conditions in running analysis.
Various IMU-based methods have been proposed in the literature based on different positions, signal types, and processing algorithms.
The authors state that 'a comprehensive guideline recommending the most suitable method for a specific experimental condition is still lacking.'
The study aimed to fill this gap by providing practical recommendations to guide method selection for IMU-based temporal parameter estimation in running.
The findings are intended to support both performance assessment and injury risk prevention applications.
Conclusions
The study provides practical recommendations for IMU-based method selection depending on device position and running speed for temporal parameter estimation.
Results indicated a clear hierarchy of device position accuracy: foot > shank > lower back.
Speed-dependent performance differences were identified specifically for stance phase estimation.
Recommendations are framed for 'in-field' use, emphasizing ecological validity for sports applications.
The findings are meant to guide researchers and practitioners in selecting appropriate algorithms given their specific dataset conditions.
What This Means
This research suggests that where you place a motion sensor (IMU) on the body and how fast someone is running both matter significantly when trying to accurately measure running timing parameters like how long a foot stays on the ground (stance phase) and how long each full stride takes. The study tested 36 different mathematical methods for calculating these timing parameters using sensors placed at the foot, shin, and lower back in 10 runners moving at three different speeds (roughly 9, 12, and 15 miles per hour). Sensors attached to the foot were the most accurate, with errors as small as 0 milliseconds on average, while sensors on the shin and lower back were progressively less accurate.
An important practical finding is that running speed specifically affects how accurately sensors can measure the stance phase (time the foot is on the ground), meaning that a method that works well at slower speeds may not work as well at faster speeds. This matters because many studies use a single algorithm regardless of the speed conditions being studied. The researchers recommend that both the sensor placement and the runner's speed be considered together when choosing which method to use for analysis.
This research matters because accurate measurement of running timing parameters is used both to evaluate athletic performance and to identify injury risk factors. By systematically comparing 36 different approaches under consistent conditions, this study offers practical guidance for sports scientists and coaches on which sensor placement and analysis method to choose for their specific situation, helping to standardize practices in a field where there was previously no clear consensus.
Check Your Own Numbers
Upload your bloodwork. We'll cross-reference your results against this study and 4,700 others.
Lubrano M, Rossanigo R, Cereatti A, Cuppini C, Fantozzi S. (2026). In-field temporal phase analysis of running: performance assessment of 36 IMU-based algorithms across different device positions and running speeds.. Journal of biomechanics. https://doi.org/10.1016/j.jbiomech.2026.113539