Wearable technologies support athletes and coaches by providing objective data to enhance performance, prevent injuries, and ptimize training. Magnetic and Inertial Measurement Units (MIMUs) measure spatio-temporal parameters, enabling the analysis of gait events outside laboratory settings. Recently, manufacturers have introduced single-point inertial sensors worn on the upper torso, gaining popularity in sports. Integrating machine learning (ML) algorithms has further revolutionized sport performance analysis. This paper presents the development and validation of a measurement method using a MIMU sensor on the upper torso to estimate gait parameters, particularly stride duration during walking and running. Data from the MIMU device is compared with that from sensorized insoles, and a metrological characterization assesses the device’s accuracy and reliability. Furthermore, the study developed ML models to classify the athlete’s activity level at three intensities: resting, walking (0 < speed ≤ 5 km/h), and running (speed > 5 km/h). Results show that the wearable sensor is accurate (bias: -0.002 s and almost 0 s for left and right feet, respectively) and precise (standard deviation of residuals: 0.07 s and 0.05 s for left and right feet) in estimating stride duration. Bland-Altman plots confirm agreement with the reference device, and Pearson’s correlation coefficients indicate strong linear correlation (0.94 and 0.97 for left and right feet). The ML-based classifier achieves an overall accuracy of 78%. Thus, the MIMU sensor placed at the upper torso effectively identifies stride duration, and the combination with the ML classifier advances wearable activity monitoring.These findings highlight the practical applicability of the proposed approach for continuous, non-invasive monitoring of athletes’ training load and movement patterns, offering a portable solution for performance optimization and injury prevention.
Measurement of Spatio-Temporal Gait Parameters through a Wearable Device for the Evaluation of the Activity Level of Athletes
Gloria Cosoli
;Marco Arnesano;Lorenzo Scalise
2025-01-01
Abstract
Wearable technologies support athletes and coaches by providing objective data to enhance performance, prevent injuries, and ptimize training. Magnetic and Inertial Measurement Units (MIMUs) measure spatio-temporal parameters, enabling the analysis of gait events outside laboratory settings. Recently, manufacturers have introduced single-point inertial sensors worn on the upper torso, gaining popularity in sports. Integrating machine learning (ML) algorithms has further revolutionized sport performance analysis. This paper presents the development and validation of a measurement method using a MIMU sensor on the upper torso to estimate gait parameters, particularly stride duration during walking and running. Data from the MIMU device is compared with that from sensorized insoles, and a metrological characterization assesses the device’s accuracy and reliability. Furthermore, the study developed ML models to classify the athlete’s activity level at three intensities: resting, walking (0 < speed ≤ 5 km/h), and running (speed > 5 km/h). Results show that the wearable sensor is accurate (bias: -0.002 s and almost 0 s for left and right feet, respectively) and precise (standard deviation of residuals: 0.07 s and 0.05 s for left and right feet) in estimating stride duration. Bland-Altman plots confirm agreement with the reference device, and Pearson’s correlation coefficients indicate strong linear correlation (0.94 and 0.97 for left and right feet). The ML-based classifier achieves an overall accuracy of 78%. Thus, the MIMU sensor placed at the upper torso effectively identifies stride duration, and the combination with the ML classifier advances wearable activity monitoring.These findings highlight the practical applicability of the proposed approach for continuous, non-invasive monitoring of athletes’ training load and movement patterns, offering a portable solution for performance optimization and injury prevention.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
