Skip to main content
. 2018 Mar 22;9:218. doi: 10.3389/fphys.2018.00218

Table 2.

Accuracy of estimated knee flexion/extension (F/E) angles (using ANN1) with different training outputs (namely: IMU or Plug-in Gait-based), using single subject training and evaluation.

Subjects IMU Plug-in Gait
Left F/E Right F/E Left F/E Right F/E
ρ RMSE (σ) ρ RMSE (σ) ρ RMSE (σ) ρ RMSE (σ)
S01 0.99 3.24 (1.53) 0.99 4.38 (1.71) 0.99 3.56 (0.97) 0.99 4.76 (1.46)
S02 0.99 1.74 (0.48) 0.99 1.77 (0.54) 0.99 4.14 (1.39) 0.99 3.79 (1.41)
S03 0.99 2.65 (0.64) 0.99 2.05 (0.53) 0.99 3.70 (1.22) 0.99 2.58 (0.72)
S04 0.99 2.60 (0.47) 0.99 2.26 (0.58) 0.99 3.02 (1.28) 0.99 3.59 (1.41)
S05 0.99 3.39 (1.79) 0.99 3.55 (2.05) 0.99 4.03 (1.19) 0.99 4.49 (1.33)
S06 0.99 3.57 (0.67) 0.99 3.52 (0.64) 0.99 2.62 (0.54) 0.99 2.27 (0.63)
S07 0.99 3.30 (0.57) 0.99 2.86 (0.51) 0.99 5.27 (1.14) 0.99 5.41 (1.21)
S08 0.99 3.95 (1.70) 0.99 3.17 (1.49) 0.98 7.33 (2.68) 0.98 8.41 (3.02)

Pearson's correlation coefficient (ρ) is calculated for each stride and averaged over approximately 200 strides for each subject (S01, S02, S03, S04, S05, S06, S07, and S08). The Root Mean Squared Error (RMSE) is calculated similarly over all strides. Training of the artificial neural networks was performed using running data at 10 and 14 km/h, while 12 km/h running data was used for evaluation.