Biomedical Engineering Reference
In-Depth Information
Fig. 5 The estimated
x
and
y
coordinates of knee and ankle motions in anterior and lateral views.
(horizontal axis: frame number, vertical axis:
x
or
y
displacement)
Table 1 Average estimation errors for knee and ankle (in pixels)
Anterior view
Lateral view
Knee
Ankle
Knee
Ankle
Right
Left
Right
Left
Mean
3.0
2.7
1.4
1.2
2.9
2.1
Std
2.7
1.8
1.1
1.1
1.4
1.5
In order to quantitatively evaluate the accuracy of the proposed approach,
we manually marked knees and ankles, which are easier to identify than hips, in
five different sequences. Two sequences had a clean background, two were with
cluttered background, and the other one with low quality video downloaded from
Youtube at [
14
]. The average estimation errors for the knee and ankle in anterior
and lateral views are summarized in Table
1
. An example of the framewise
estimation errors of the right knee and ankle is given in Fig.
6
.
6 Conclusions
In this paper, we have shown an effective approach to estimate human body joints
by incorporating a skeleton-based graph growing method and loose stick figure
models. No markers are needed, and initialization is easy and flexible in the
proposed framework. The results of the experiments demonstrate the robustness
and accuracy of the proposed algorithm in key joint estimation and jumping
analysis. We are collecting clinical data to evaluate joint angles using the method
