Skip to main content

Table 2 Average (standard deviation) of the gait variables calculated for the motion capture and video data using the three pose tracking algorithms

From: Concurrent validity of human pose tracking in video for measuring gait parameters in older adults: a preliminary analysis with multiple trackers, viewing angles, and walking directions

Gait variable

Motion capture

AlphaPose

OpenPose

Detectron

Front view walks

Back view walks

Front view walks

Back view walks

Front view walks

Back view walks

Front view walks

Back view walks

Top camera

Eye-level camera

Top camera

Eye-level camera

Top camera

Eye-level camera

Top camera

Eye-level camera

Top camera

Eye-level camera

Top camera

Eye-level camera

Cadence (steps per minute)

106.44 (8.10)

105.30 (8.98)

106.23 (9.14)

106.51 (8.62)

105.63 (9.39)

105.26 (9.30)

106.22 (9.14)

106.51 (8.62)

105.63 (9.40)

105.26 (9.29)

106.23 (9.14)

106.52 (8.62)

105.63 (9.39)

105.26 (9.29)

Step time (s)

0.57 (0.06)

0.58 (0.07)

0.57 (0.10)

0.57 (0.06)

0.58 (0.17)

0.58 (0.07)

0.57 (0.10)

0.57 (0.06)

0.58 (0.17)

0.58 (0.07)

0.57 (0.10)

0.57 (0.06)

0.58 (0.17)

0.58 (0.07)

Step width (m)*

0.09 (0.05)

0.10 (0.04)

0.63 (0.20)

0.62 (0.20)

0.55 (0.16)

0.54 (0.16)

0.62 (0.46)

0.67 (0.29)

0.59 (0.23)

0.54 (0.22)

0.66 (0.19)

0.64 (0.19)

0.86 (1.92)

0.54 (0.15)

CV step time

0.06 (0.03)

0.06 (0.04)

0.13 (0.10)

0.06 (0.02)

0.16 (0.18)

0.06 (0.03)

0.13 (0.10)

0.06 (0.02)

0.16 (0.18)

0.06 (0.03)

0.13 (0.10)

0.06 (0.02)

0.16 (0.18)

0.06 (0.03)

CV step width

0.42 (0.21)

0.37 (0.16)

0.27 (0.08)

0.30 (0.07)

0.27 (0.08)

0.28 (0.07)

0.38 (0.21)

0.38 (0.12)

0.37 (0.10)

0.39 (0.09)

0.25 (0.07)

0.28 (0.06)

0.29 (0.12)

0.27 (0.08)

eMOS (m)a

0.04 (0.02)

0.04 (0.02)

0.25 (0.12)

0.23 (0.10)

0.22 (0.09)

0.21 (0.08)

0.33 (0.14)

0.30 (0.14)

0.28 (0.12)

0.25 (0.11)

0.25 (0.12)

0.23 (0.09)

0.40 (0.15)

0.21 (0.09)

  1. aNote that except for cadence and step time, the values calculated from the pose tracking algorithms are dimensionless (normalized by hip width) and cannot be directly compared with motion capture values