Information Technology Reference
In-Depth Information
plane
1
and the participant's height
2
. For both systems, participants were given a
practice trial in order to familiarize with the mechanics of the test and to confirm the
fully understanding of the instructions. Later, they were instructed to complete a full
CSRT task where execution of the test was recorded by the corresponding controlling
software. Once the test was completed, they were asked to complete a practice trial
and a full CSRT task on the other platform. In this experiment, the software that
controls the MAT and the Kinect were run in the same computer to minimize bias in
the measurements due to variability in computational power. As instructed in the
original version of the CSRT test, the mean Reaction Time (RT) of 20 trials is
measured and analyzed. Also, the mean Decision Time (DT) and Movement Time
(MT) were examined for a more descriptive assessment.
In order to determine the validity of the Kinect-based CSRT, this evaluation was
conducted in two steps. Firstly, the consistency and association of the three
parameters (DT, MT and RT) per participant per panel was assessed. Shapiro-Wilk
test was utilized to assess the distribution of the time-based variables. The consistency
and association of the three parameters (DT, MT and RT) were assessed by
calculating the Pearson Correlation Coefficient and the Intraclass Correlation
Coefficient model 3,1. Secondly, the RT values were used to compute the CSRT test
for both the Kinect and MAT and a similar analysis was then conducted.
4
Results
Regarding to the consistency of the Reaction Times (RTs), the Shapiro-Wilk test showed
normality across the log transformed time-based measures of the CSRT (Sig. > 0.05).
The Pearson Correlation Coefficient showed high association between the values
collected by Kinect and the MAT (Pearson r = .746) (Refer to Table 1). Likewise, the
ICC model 3,1 confirmed the consistency and agreement of the RT values across the
samples (ICC Single Measures = .657, Average Measures= .793, Sig = .000).
Regarding the consistency of the Decision Times (DT) and Movement Times
(MT), complementary parameters computed on the MAT version of the CSRT test;
the Pearson r analysis determined that the DT values were highly correlated; however,
the MT values reported a poor association. (Refer to Table 1). Finally, the Choice
Stepping Reaction Time (CSRT) test was calculated for both the Kinect and MAT as
instructed in its original version. The Pearson r analysis showed high association
between the CSRT values obtained by the Kinect and the MAT (Refer to Table 2).
Similarly, the ICC model 3,1 confirmed the consistency and agreement between the
CSRT measures computed with the Kinect and the MAT. (ICC Single Measures =
.645, Average Measures= .784, Sig .016).
1
To minimize bias in data collection due uneven surfaces or placing the Kinect on a slightly
tilted table, a short calibration sequence was incorporated to determine the orientation of the
actual floor plane (or ground).
2
The participants' height is also measures and used to determine the distance of the panels (in
this paper known as Expected Step Length). This makes the test equally challenging for
audiences as panels are dynamically located based on the user's height which also takes away
the limitation of the original test where the step panels are fixed.
