Fig. 2
From: Novel velocity estimation for symmetric and asymmetric self-paced treadmill training
Velocity estimation error during fixed-speed treadmill walking for the three tested velocities. a Mean estimated velocities across subjects for several kinematic and kinetic algorithms using 100 steps on the treadmill in fixed-speed mode. b Distribution plots of the velocity estimation error for each of the algorithms. The impulse-momentum algorithm is not shown because it accumulates significant error. Using the 1-segment push-off correction to step length (PO-1S) resulted in a statistically significant reduction in the velocity estimation error at all three normalized velocities with respect to the no push-off model (No PO) (Signed-Rank Test: Froude Number = 0.075, p = 0.0016; Froude Number = 0.150, p = 0.0078; Froude Number = 0.225, p = 0.0078). We did not observe any significant differences between the 1-segment model (PO-1S) and the 2-segment model (PO-2S). The 1-segment model (PO-1S) resulted in a statistically significant reduction of error at the lowest and fastest speeds with respect to the leg-swing method (LS) (Signed-Rank Test: Froude Number = 0.075, p = 0.0078; Froude Number = 0.150, p = 0.0547; Froude Number = 0.225, p = 0.0156)