*3.1. Lab Study*

Table 4 depicts the mean errors and standard deviations for both stride velocity and stride length of the four different algorithms for the lab study dataset. The results were averaged over all strides in the lab study dataset. The results show that the *Trajectory* algorithm performed best considering both the ME ± Std and the MAE.


**Table 4.** Mean error (ME) and standard deviations (Std), mean percentage error (MAPE), and mean absolute error (MAE) of stride length and average velocity per stride of the four algorithms for the lab study dataset.

Figure 7 shows the results of the stride length for the different velocity ranges. The MEs of the *Deep Learning* algorithm increased with higher velocities. The *Trajectory* showed lower MEs for the three slower velocity ranges than for the highest velocity range. The *Acceleration* algorithm showed small errors from 3–5 m/s. Its performance dropped for the outer velocity ranges from 2–3 m/s and from 5–6 m/s. The *Stride time* algorithm worked well for the velocity range of 2–3 m/s and 5–6 m/s; however, it showed large errors of more than 40 cm for the other velocity ranges.

**Figure 7.** Mean error of the stride length of the four different algorithms for the different velocity ranges the subjects ran in the lab study.

Figure 8 shows the Bland–Altman plots for both the stride length and the average velocity per stride for the lab study dataset. The results are color coded into the velocity ranges presented in Table 2. The *Trajectory* algorithm performed well for velocities up to 5 m/s. For the high velocity range, larger errors could be observed. The *Stride time* algorithm performed worst and showed a linear error distribution in the Bland–Altman plots. In the *Acceleration*, *Trajectory*, and *Deep Learning* plots for stride length, we see the samples of the different velocity ranges overlapping. This overlap is not visible in the velocity plots.

(**a**) Algorithm: *Stride time*; metric: stride length (**b**) Algorithm: *Stride time*; metric: velocity

(**c**) Algorithm: *Acceleration*; metric: stride length (**d**) Algorithm: *Acceleration*; metric: velocity

(**e**) Algorithm: *Trajectory*; metric: stride length (**f**) Algorithm: *Trajectory*; metric: velocity

(**g**) Algorithm: *Deep Learning*; metric: stride length (**h**) Algorithm: *Deep Learning*; metric: velocity

**Figure 8.** Bland–Altman plots for stride length and velocity for the four algorithms. Each row contains the metrics for one algorithm. The individual samples are color coded depending on the velocity bin of the sample: 2–3 m/s blue, 3–4 m/s red, 4–5 m/s green, 5–6 m/s purple. The dotted-dashed horizontal lines depict the mean error and the dotted horizontal line the 95% confidence interval.
