Professional Documents
Culture Documents
Gait Report
Gait Report
Abstract:
This gait analysis investigated the individual walking pattern of each of our group
members. Through the study of one’s acceleration, velocity, arm movement, and frequency
of steps over a given time period, we were able to compile an abundance of data in order
to break down the simple act of walking. We found that all group members had a
consistent stride but variation in acceleration.
Table of Contents
2
Introduction
The purpose of this experiment was to analyze the simple action of walking. Because
everyone walks differently, we wanted to create an equation that determines the
acceleration of someone’s lower back in three directions.
We conducted this experiment to learn how to use data from multiple people to create
consistent equations that determine certain output values. For this project, we analyzed the
acceleration from 4 different people to form equations that find acceleration in x, y, and z
directions based on time. We also created an equation to predict someone’s average
absolute (always positive) acceleration based on their height.
Methods
We conducted two trials for each of our four test subjects, gathering data on acceleration
and velocity in the x, y, and z directions, as well as the absolute acceleration and velocity,
the total number of steps, the total time elapsed, the average arm extension, and the
height of the subjects.
We chose to only conduct trials for four test subjects because we had enough diversity in
our group. Not only did we have people of different heights, ranging from 65 to 72.5 in, but
we also had both male and female test subjects.
We installed the app “Phyphox” on an android phone to collect the acceleration data. We
chose the option without gravity because adding gravity would have made the data
unnecessarily more difficult to analyze. We attached the phone to the test subjects’ lower
backs using duct tape, and had the subjects walk a distance of 6 meters. One person not
being tested counted the steps of the test subject and another made sure to stop the
accelerometer when the subject reached the 6 m mark. The data we collected is saved on a
Google Sheets where the data is converted to several visual graphs. We also measured the
average arm extension by laying down a tape measure and measuring the farthest point to
which the subject extends his/her arm back and forward and calculated the difference.
3
We used the acceleration to calculate the velocity, which can be found on our Google
Sheets. We calculated this by using the midpoint sum formula, which is a way of
approximating an integral (the area under the acceleration vs time graph). The formula is
(t2 - t1) * average (a2, a1) + a1, where t1 and
t2 are two different times, and a1 and a2 are
the
two different accelerations at t1 and t2. We chose not to include these velocity graphs on
our report because they were inconclusive. The results suggested were the same as the
results suggested by the acceleration graphs, so they were unnecessary to include.
However, they can be seen at the links found under Results and Discussion.
Below are the acceleration results for Rohan Ayyar. His data can be found at L
INK. He is
68.5 inches tall, took an average of 11.5 steps over his two trials and had an average arm
extension length of 22 inches.
These results suggest that Rohan does not move his torso much when walking; doing so
would likely result in a higher acceleration on the lateral axis (x). When this data is
compared to the acceleration in the x-direction of everyone else, it is evident that Rohan’s
data varies less on average than everyone else. However, there is variation. This may be
because the phone may have not been fully secure on Rohan’s back.
4
The results for y-acceleration suggest that Rohan has a consistent stride regarding his
vertical movement. The times of the steps are clear and evenly divided, suggesting a
consistent walk. Because this data shows Rohan’s vertical acceleration, it can be concluded
that Rohan lifts himself up and down in a similar manner for every step.
The results for the z-direction suggest that Rohan slows down and speeds up consistently
with each step. Trial 1 shows Rohan reaching about the same acceleration in the z-direction
for every step, close to around 3.5 m/s^2. Trial 2 shows a similar pattern but the
acceleration goes all the way up to 6 m/s^2 for some steps. This suggests that Rohan took
more aggressive strides in his second trial.
These results suggest that Rohan’s absolute acceleration was very scattered. His absolute
acceleration for both trials generally increases until around t = 4 seconds, then begins to
slowly decline. This makes sense because Rohan likely didn’t reach his full stride until
around 4 seconds, then began to slow down as he approached the 6 m mark.
Next up is the data for Nick N. His data can be found at LINK. He is 65 inches tall, took an
average of 10 steps over his two trials and had an average arm extension length of 28
inches.
5
The results for x-acceleration suggest that the Nick N moves his torso quite a bit when
walking, especially in Trial 2. In Trial 1, the numbers are about the same as Rohan’s
acceleration and velocity, but in trial 2, either Nick N walked much more aggressively, or the
phone got loose from the tape. This would have resulted in a few jerks as the phone would
be much more prone to move.
These results suggest that Nick N is consistent with step timing, but not with the force with
which he lifts himself up when walking. As seen in the graphs, there are a few spikes that
almost reach to 15 m/s^2, likely caused by Nick N taking hard steps.
6
The results for the z-direction suggest that Nick N was consistent for Trial 1, but not for
Trial 2. In Trial 1, Nick N takes relatively even steps with relatively equal magnitudes. In Trial
2, however, Nick N has a few strides that reach 20 m/s^2 in the z-direction. This suggests
that in Trial 2, Nick N took a few long, aggressive strides that resulted in high accelerations.
The results for absolute acceleration corroborate our data for the z-direction. In Trial 2,
Nick N had a few steps that were more aggressive than the others.
Next up is the data for Nick B (Nic). His data can be found at L
INK. He is 72.5 inches tall,
took an average of 10 steps over his two trials and had an average arm extension length of
28 inches.
7
The results for the x-direction indicate that either Nick B moved his torso aggressively in
both trials, or the phone was insecure and moved around for both trials. It is important to
note that Trial 2 has less variation, but both still have a lot.
The results for acceleration in the y-direction suggest that Nick B mostly took even strides.
However, in the beginning and towards the end of the time interval, the acceleration looks
to be close to 0 m/s^2.
8
These results suggest that Nick B walked in a normal, even manner for Trial 1, but walked
more aggressively for Trial 2 (on certain steps). This could suggest that Nick B took a few
long strides in Trial 2 to account for those spikes, or that the phone was insecure. Given
that some of our other data is similarly random, the phone was likely not fully secure.
These results suggest that Nick B walked normally for Trial 1, other than the odd spike in
the beginning (probably caused when the start button was pushed and the phone moved).
Nick B sped up to a normal stride, then began to slow down. Trial 2 suggests Nick B kept
the same stride for most of the duration and spent less time speeding up and slowing
down.
9
The results for the x-direction suggest that Kelsey moves her torso fairly aggressively, as
indicated by the variation in the velocity and acceleration graphs. This also suggests that
Kelsey reached the peak of her stride at around t=5 seconds, because the values for x
acceleration and velocity are the highest. Generally, people move their torso more when
they walk faster, and these graphs support that statement.
The results for acceleration and velocity in the y-direction show how Kelsey lifts herself up
when walking. Trial 1 is interesting because she appears to stop lifting herself up as much
at around t=3 seconds, then accelerates in the y-direction for the next two steps, then
begins to decelerate. In Trial 2, however, Kelsey is much more consistent (until t=6
seconds). She forced herself up and down at a similar rate for each step.
These results suggest that Kelsey speeds up and slows down consistently with each step in
the z-direction. She consistently reaches around 4 m/s^2, which, although more than Nick B
and less than Nick N on average, is pretty close to Rohan. The spike at around t=7.5 s could
be an aggressive stride forward, or the more likely reason, an insecurity of the phone on
Kelsey’s back.
10
These results suggest that overall, Kelsey walked somewhat consistently. Clearly Trial 2 is
more consistent, but both trials show a similar trend: Kelsey’s absolute acceleration
increases with each step, then plateaus, then decreases. This boosts the credibility of the
data because it suggests she sped up to her normal stride, then kept it for some time, then
began to slow down as she approached the end.
Other Data
The other data we collected includes number of steps, height, average arm extension,
frequency, and period.
11
Based on this data, there appears to be no correlation between height and average
number of steps taken, or between height and average arm extension. Of course, we only
have data from four subjects, so it can’t be concluded that there is no correlation; we just
didn’t seem to find any. There appears to be a weak positive correlation between height
and frequency, and a weak negative correlation between height and period.
Models
We created models to predict the acceleration of the lower back in all three directions for
both males and females. Our data suggested that acceleration is approximately sinusoidal
for all three directions. We calculated these equations using the average peaks of the
acceleration graphs as the amplitude, and the average frequencies of everyone as the
frequency. Our first model predicts the acceleration in the x-direction for a 6 meter walk.
The equation, where a(t) = x acceleration (m/s^2), and t = time (s), is a(t) = 2.12sin(5.27t).
This looks like:
12
For example, at time = 5.98 s, Rohan’s actual acceleration is 0.268 m/s^2. Our model for the
x-direction predicts an acceleration of 0.2088 m/s^2. The residual for this point is 0.059
m/s^2, which is low enough to suggest accuracy. However, some points have higher
residuals.
Our second model predicts the acceleration in the y-direction for a 6 meter walk. The
equation, where a(t) = y acceleration (m/s^2), and t = time (s), is a(t) = 5.55sin(5.27t). This
looks like:
Our third and final model predicts the acceleration in the z-direction for a 6 meter walk. The
equation, where a(t) = y acceleration (m/s^2), and t = time (s), is a(t) = 5.308sin(5.27t). This
looks like:
13
These models can be used as a rough predictor of acceleration in a certain direction over a
given time interval. From the models we can conclude that people generally accelerate their
lower back in the direction they are walking and vertically more than horizontally.
We would have included a separate model for females, but our data for females and males
were very similar, so we decided against it. These models would also be more accurate if
we had taken more tests.
Another model we created models the height and predicted average absolute acceleration
for a 6 m walk. The regression line equation is a = 8.81 - 0.098h, where h is the height (in)
and a is the average absolute acceleration (m/s^2).
There appears to be a moderate negative correlation between height (in) and average
absolute acceleration (m/s^2). Thus, a tall person is more likely to have a lower average
absolute acceleration, and a short person is more likely to have a higher absolute
acceleration. Someone with a height of 70 in is predicted to have an average absolute
acceleration over a 6 m walk of about 1.97 m/s^2 (found by plugging 70 into the equation
for h and solving for a). Obviously, this model isn’t 100% accurate; the correlation is only
14
-0.67, which is fairly strong, but not enough to make any conclusions. Also, there are
possibilities of error in the testing itself, in which the phone has been assumed to not be
fully secure. In addition, we only have four data points, which isn’t really enough to
accurately predict the acceleration. If we had more time, we would have conducted tests
for more people with different heights and of both genders, and used that data to make
more accurate and predictive models. For now, however, these models suffice.
References
https://youtu.be/5j4YRHf6Iyo
https://docs.google.com/spreadsheets/d/11qEGnXsaoH1GvTlnnwn5vnwx5e2IjgjK/edit#gid=
1008726577
https://docs.google.com/spreadsheets/d/12PUIXWguD7xTXMgvcmTE71LM12gl_XRZ/edit#gi
d=1812600558
https://docs.google.com/spreadsheets/d/13aOhO8hfEP5bag2iv888E8sqwW3PZuZX/edit#gi
d=839192430
https://docs.google.com/spreadsheets/d/14-ymJ4EYrJnMIR5hj0U4cYs4YrJr3lQi/edit#gid=110
0071029
Appendix
15