**Section Side Kinetic Detection Acceleration Detection Integrated Detection** *F p* **Mean (SD), ms Mean (SD), ms ∆% LOA% Mean (SD), ms ∆% LOA% 4. Discussion**

ICC: intra-class coefficient; 95% CI: 95% confidence interval.

Straight Right 713.1 (243.3) 730.5 (252.2)\* 2.4 95.4 738.8 (259.4)\* 3.6 94.2 15.236 <0.001 Left 736.7 (261.2) 740.2 (250.1) 0.5 91.8 744.8 (264.6) 1.1 90.1 0.670 0.512 Curve Right 614.7 (142.6) 629.4 (153.5)\* 2.4 96.1 632.3 (150.2) \* 2.9 93.4 92.298 <0.001 Left 587.6 (127.1) 587.8 (108.5) 0.0 93.8 583.3 (102.2) 0.7 95.0 0.479 0.619 The differences between the acceleration and integrated detection methods and the kinetic detection methods are shown as ∆%. In this study, we aimed to estimate the accuracy of IMUs for the phase identification of long-track speed skating for competitive speed skaters by comparing it with phase identification using the foot pressure sensor system. We examined the agreements of the acceleration and integrated detection methods with the gold standard measurements (i.e., the kinetic detection method) to calculate the stance time based on the foot contact and foot-off identified by each detection method.

The proportion of cases within the limits of agreement is shown as LOA%. The *F* value and *p* value were obtained by the repeated measures analysis of variance. \* Significantly different from the kinetic detection method in the post-hoc analysis. **Table 5.** The intra-class coefficient as computed by the acceleration and integrated detection methods. **Section Detection Method Right Left**  ICC (2,1) [95% CI] ICC (2,1) [95% CI] Straight Acceleration 0.927 [0.906−0.943] 0.882 [0.852−0.907] Integrated 0.948 [0.925−0.963] 0.868 [0.834−0.895] Curve Acceleration 0.904 [0.875−0.926] 0.657 [0.582−0.721] Integrated 0.891 [0.529−0.956] 0.700 [0.633−0.757] The main finding of this study is the high degree of agreement between the kinetic and acceleration/integrated detection methods measured with the foot pressure sensor and IMU systems, as shown by the moderate to high ICCs. This was true for both sides (left and right) and segments (the straight and the curve). While statistically significant differences between the methods were found for the stroke time on the right side for both the straight and the curve, these differences were within 3.6%. The significant difference may partly be due to the large number of strokes used for the comparison, while the magnitude of the observed errors may not be very meaningful. Our Bland-Altman analysis shows that in the straight, the extent of the bias was proportional to the observed stance time (Figure 2). It is known that, during running, the stance time is prolonged as the running speed decreases [19]. Therefore, it should be noted that both the acceleration and integrated methods may be biased when the stance time is greater and the skating

speed is slower (e.g., during long-distance skating). Our results also suggest that during the curve, the ICCs for the left side were substantially lower than those for the right side. This side-specific difference may be related to the asymmetrical skating form during the curve. Further study is needed to investigate the side-specific difference of the skating form during the curve in speed skating.

The acceleration and integrated detection methods were in significant agreement with the gold standard measure for the computation of the stance time, which suggests that the timing of the foot contact and foot-off for each leg and stroke can be accurately detected by these methods. The identification of foot contact and foot-off during skating is crucial for characterizing skating performance. It has been shown that the force measured by the sensor embedded in the skate shoe is greater when the subject stands on one leg (single leg stance), while the force substantially decreases when both legs are on the ice (double leg stance) [20]. The detection methods proposed in this study can be used to characterize skating performance using only IMUs, with minimal interference to the performance of the subject. The accuracy of acceleration and integrated methods was similar in our study, suggesting either method can be used for the detection of foot contact and foot-off. However, the phases of the speed skating motion can be divided into more details than just foot contact and foot-off [21]. IMUs have the potential to be used to identify a more detailed phase classification. In particular, the knee flexion, hip flexion and hip extension angles may potentially be used for a more detailed phase classification, as these angles show phase-dependent changes [22]. Therefore, the combined use of acceleration and the joint angle profiles obtained by IMUs would be ideal for future studies.

This study had several limitations. First, we only included healthy competitive athletes from a university long-track speed skating team. Further studies are necessary to generalize the results to different populations. Second, we used a foot pressure system as the gold standard measure, although the system itself could exhibit a measurement bias. Specifically, we used 20% peak force as the threshold for the foot contact and foot-off timing for the kinetic detection method. The 20% threshold was selected based on the observation of all trials from all participants, assuring no false detection in the kinetic detection, while the threshold may not be generalizable to other datasets. Furthermore, in reality, foot contact and foot-off occurred respectively earlier and later than the timing identified by the kinetic detection. This time lag between the actual and detected events can explain the systematic bias observed between the detection methods (i.e., all the positive ∆% values in Table 4). This time lag can overestimate the bias, while providing conservative results for the objective of this study.

#### **5. Conclusions**

In this study, we examined the agreement among the acceleration and integrated detection methods and the gold standard measure (i.e., the kinetic detection method) to calculate the stance time based on the foot contact and foot-off identified by each detection method. Despite the statistically significant differences between the acceleration/integrated detection methods and the gold standard measure on the right side, these differences were within 3.6%. The current data show that phase identification using acceleration and integrated detection is valid for evaluating the kinematic characteristics during long-track speed skating.

**Author Contributions:** Conceptualization, Y.T., T.I., K.I. and S.I.; methodology, Y.T. and T.I.; data collection, Y.T. and T.I.; formal analysis, Y.T. and T.I.; data curation, Y.T. and T.I.; writing—original draft preparation, Y.T. and T.I.; writing—review and editing, Y.T., T.I., K.I. and S.I.; supervision, Y.T. and S.I.; funding acquisition, Y.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was supported by Kakenhi (Grant-in-Aid for Early-Career Scientists No. 19K20011).

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of the Takasaki University of Health and Welfare (approval number: 1904, 17 May 2019).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** Data available on request due to ethical restrictions.

**Acknowledgments:** We thank all participants who volunteered for our study.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Sensors* Editorial Office E-mail: sensors@mdpi.com www.mdpi.com/journal/sensors

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34

www.mdpi.com

ISBN 978-3-0365-5859-2