*2.4. Statistical Analysis*

α Means and standard deviations of the differences in kinematic parameters were estimated based on individual measurements between systems. All dependent variables were assessed for normality using a one-sample Kolmogorov–Smirnov test (K-S test, α = 0.05). A two-tailed paired *t*-test was employed based on the normally distributed data to test the differences between the two systems. The effect size was assessed using Cohen's d [30]. An alpha level of 0.05 was used for statistical analysis. SPSS (22.0, IBM Inc.; Chicago, IL, USA) was used to conduct all statistical analyses. The alpha level was adjusted by 30 dependent variables using the Bonferroni correction to reduce the chances of type I error (α = 0.05/30 = 0.0017).

#### **3. Results**

α

Sixteen participants (9 males and 7 females) participated. The participants' ages, body mass, and height were 23.44 ± 2.31 years, 69.72 ± 9.82 kg, and 1.73 ± 0.08 m, respectively.

#### *3.1. Lower Extremity Joint Moments and Powers*

Ensemble curves of lower extremity sagittal plane moments and powers estimated using MB and ML are presented in Figures 1 and 2, respectively. Scaled (by body mass) peak magnitudes and relative timing to the peak are presented in Tables 1 and 2, respectively.

Paired *t*-tests used for analysis based the results that confirmed the normality of the outcome variables. Compared to the MB system, the ML system showed significantly greater peak joint moment magnitudes at HM<sup>2</sup> (ML: −1.73 ± 0.27, MB: −1.38 ± 0.29), HM<sup>3</sup> (ML: 2.27 ± 0.45, MB: 1.42 ± 0.29), KM<sup>2</sup> (ML: −1.17 ± 0.24, MB: −0.74 ± 0.13), and AM<sup>1</sup> (ML: 3.32 ± 0.55, MB: 3.14 ± 0.51), but less peak magnitude at KM<sup>1</sup> (ML: 1.28 ± 0.32, MB: 1.40 ± 0.42). For the joint powers, significantly less peak magnitudes were at KP<sup>1</sup> (ML: −4.05 ± 1.79, MB: −5.0 ± 2.77), KP<sup>2</sup> (ML: 2.64 ± 1.09, MB: 3.15 ± 1.41), but were greater at HP<sup>2</sup> (ML: 8.07 ± 2.11, MB: 4.29 ± 1.14), HP<sup>3</sup> (ML: 5.68 ± 2.71, MB: 3.99 ± 2.13), KP<sup>3</sup> (ML: −5.42 ± 1.61, MB: −3.45 ± 1.29), KP<sup>4</sup> (ML: −9.65 ± 2.10, MB: −7.15 ± 1.83), AP<sup>1</sup> (ML: −9.44 ± 1.81, MB: −8.38 ± 2.48), as well as AP<sup>2</sup> (ML: 18.40 ± 4.91, MB: 16.07 ± 3.60). In addition, the relative timing to the peak was detected to be significantly different between the MB and ML systems. To be specific, the ML system took longer than the MB system to reach the HM<sup>1</sup> (ML: 6.74 ± 3.40, MB: 5.16 ± 1.27), HM<sup>2</sup> (ML: 43.59 ± 7.00, MB: 40.26 ± 6.90), HM<sup>3</sup> (ML: 92.73 ± 3.00, MB: 90.58 ± 3.39), KM<sup>1</sup> (ML: 13.53 ± 3.89, MB: 12.98 ± 2.18), KM<sup>2</sup> (ML: 92.19 ± 2.51, MB: 90.93 ± 2.21), HP<sup>1</sup> (ML: 28.30 ± 3.68, MB: 26.16 ± 4.45), and KP<sup>1</sup> (ML: 8.73 ± 3.01, MB: 8.19 ± 2.21). Besides, the ML system took less time than the MB system to reach the HP<sup>3</sup> (ML: 89.63 ± 4.15, MB: 91.23 ± 3.47). See Tables 1 and 2 for more details.

**Table 1.** Body mass scaled peak magnitude (Mean, SD) for joint moments and powers for the marker-based (MB) and Markerless (ML) systems.


K-S tests results for peak moments and peak powers at hip, knee, and ankle joint listed here were all greater than 0.05, therefore normal distribution of the parameters listed in this table was confirmed. \* Indicates significant difference; **Bold** indicates the event was observed within the stance phase; *italic with underline* indicates the event was observed during the stance–swing transition; the rest of the parameters were observed within the swing phase. For joint moment data, "+" represents the extension (ankle plantar flexion) and "−" represents the flexion (ankle dorsiflexion). For joint power data, "+" represents energy production, and "−" represents energy absorption.


**Table 2.** Relative Time to Peak as Percentage Stride Cycle (Mean, SD) for Joint moments and Powers for Marker-based (MB) and Markerless (ML) Systems.

K-S tests results for relative timing to peak moments and peak powers at hip, knee, and ankle joints listed here were all greater than 0.05, therefore normal distribution of the parameters listed in this table was confirmed. \* Indicates significant difference; **Bold** indicates the event was observed within the stance phase; *italic with underline* indicates the event was observed during the stance–swing transition; rest of the parameters were observed within the swing phase.

#### *3.2. Lower Extremity Joint Center and Segment Center of Mass*

Joint moments could be significantly affected by joint center position and the segment center of mass. Figures 3 and 4 demonstrate the ensemble curves of differences in joint center positions (hip, knee, ankle) and segment center of mass (thigh, leg, foot) between MB and ML.

In the mediolateral direction (Figure 3 top panel), the ankle (left) and knee (middle) joint centers were biased toward the lateral direction in the ML than the MB throughout the stride cycle. The hip joint center showed the same trend except during initial contact and the late swing phase. In the anterior–posterior direction (Figure 3 middle panel), ML showed a posterior-biased hip joint center during the stride cycle, whereas the ankle and knee joint centers varied within the stride cycle. The ML was posteriorly biased compared to the MB at initial contact for both the ankle and knee joints. For the rest of the stance phase, the ML for the ankle joint was slightly more posterior, and the knee was more anterior. While the ML for the ankle joint continued in the posterior direction in the swing phase, the ML knee continued in the anterior direction in the early swing phase but turned to the posterior direction for the rest of the swing phase. In the vertical direction (Figure 3 bottom panel), the ML of the ankle varied during the stride cycle. In the early stance and mid-swing phase, the estimated bias was toward the superior direction, while in the mid-stance and early swing, it turned to the inferior direction. The ML showed an inferior-biased knee and hip joint center in the early stance. When the ML knee joint turned to the superior direction for the rest of the stride cycle, the hip joint was superior in the mid-stance and early swing phase but moved toward the inferior direction in the rest of the swing phase. See Figure 3 for more details.

**Figure 3.** Ensemble curve of lower extremity joint position differences between marker-based and markerless motion capture systems across the average 10 stride cycles for 16 participants. Differences were estimated as markerless (ML) joint center position—marker-based (MB) joint center position.

—

In the mediolateral direction (Figure 4 top panel), the center of mass of the foot, leg, and thigh was more lateral in the ML than in the MB systems throughout the stride cycle. In the anterior–posterior direction (Figure 4 middle panel), the foot center of mass was more anterior in the ML than in the MB system during the stride cycle. For the leg center of mass, ML showed more posterior biases than the MB during the stance and swing phases except in different directions at the end of the swing phase. The thigh center of mass was mainly posterior throughout the stride cycle but briefly anterior in the swing–stance transition phase. In the vertical direction, the foot center of mass showed a higher position in the ML than in the MB system during most stance and late swing phases but was lower during the stance–swing transition and early swing phase. For the leg center of mass, the ML demonstrated lower values than the MB system during about 85% of the stride cycle but they were briefly higher during the early swing phase. The thigh center of mass from the ML showed a higher position than the MB system over the whole stride cycle.

— **Figure 4.** Ensemble curve of lower extremity segment center of mass differences between markerbased and markerless motion capture systems across the average 10 stride cycles for 16 participants. Differences were estimated as markerless (ML) center of mass position—marker-based (MB) center of mass position.
