You are on page 1of 10

Gait Analysis

Tyler Colenbrander
Michelle Frid
Alexa Funahashi
Sophia Galleguillos

STEM Senior Engineering 2°


16 October 2018
Gait Analysis Table of Contents
Abstract……………………………………………………………………....……2

Introduction………………………………………………………………………..2

Method…………………………………………………………………………..2-3

Results……………………………………………………………………..…...3-6

Predictive Model………………………………………………………………..6-8

Discussion/Conclusion……………………………………………...…………...8

References………………………………………………………………....……..9

Appendix…………………………………………………………………………..9

Page 1 of 10
Abstract

For this experiment, we took data on how different people walk to determine if there are
any patterns or quirks that can help identify each person. To do this we did multiple trials of 2
subjects walking and running to find the differences between their gait, or how each person
moves. We used an app to collect data on the lateral, vertical, and anterior motions. We also
collected data on how fast each person walked and ran and the different leg lengths. We found
that the longer distance of the person’s step, the shorter the time was. For every 4 cm increased
in height, 1 cm is added to step length, which is demonstrated by Equation 1. Then for every 1
cm increase in step length, 1/20 of a second is reduced from step time. From the analysis of
these results, we can predict fairly accurately these gait parameters for walking people, but the
running tests proved to be difficult to correlate.

Introduction

This project focused on how people walk. We analyzed our group members walking
speeds and patterns to determine if there are any indicators or uniqueness about their walks.
We looked into each subject’s physical health, step length, dominant foot, and body
measurements such as height, leg length, and wingspan. Using the data and knowledge
collected from this project, we can set a baseline for certain types of peoples walking patterns
and use them to identify other people. Our group tried to pinpoint any quirks or certain ways the
subjects walked and compare them to what we considered “normal” gait. Our discovery
question is how gait frequency relates to height. In this report, we will answer the discovery
question by analyzing our data and drawing a conclusion about different gaits and how they
relate to physical characteristics like height. ​Our group came up with a unique equation for
step length: step length = 0.25(height) + 21.25. And, we came up with an equation for step
time: step time = -0.05(step length) + 3.8. This corresponds with the data that we collected
but a simple equation can be made for any subject. Scientists can use ​the data created from
or for gait analysis can be used in the law enforcement fields to help identify criminals or
unknown subjects based on their walk.

Methods

During this experiment, we took measurements of each subject that we thought were
related to one’s gait. The measurements would later help us determine the quirks of how each
person walked. We then used the “Physics Toolbox Accelerometer” app to measure the
anterior, vertical, and lateral movements of each person’s gait.
The apparatuses we used was the “Physics Toolbox Accelerometer” app, an iPhone,
measuring tape, and a stopwatch. The app senses the movement of the subjects records the
data. The iPhone was used for the app and was strapped to each subjects’ dominate leg to
accurately collect data of their gaits. The measuring tape was used to measure the length of

Page 2 of 10
each step and different body part lengths that related to their gaits. The stopwatch took the
times of the trials and helped calculate the accelerations.
For our procedure, we used our two subjects, Michelle and Alexa, and decided on the
different body measurements that would be important for determining the differences in each
person’s gait. We measured the total height of each person, the leg length, the length from knee
to foot, and foot length. We also noted each person’s dominate foot and any health problems
that may affect the way they walk or run. After collecting the body measurements, we did some
walking trials. Each person strapped the phone, with the app to collect data, on their dominate
foot. They then took 10 steps that were timed and also measured to see how long each step
was and how long it took. We did this 3 times each so we would have enough walking trials to
achieve an average. After doing the walking trials, we started our running trials. Each subject
ran 5 steps that were again times and measured. This also had 3 trials for each subject.

Results

The following table displays the general data and trends observed in our methods, which
was compiled from the raw data we collected (SEE Appendix). The trial data, which is averaged
from sets of the 12 different trials we did, shows us that the taller person has a longer step
length while walking, and this longer step length corresponds to a shorter step time (higher gait
frequency) as well. We noticed this trend and decided to base our predictive model off of it,
which is found below all the tables and graphs.

Trial Data
Step Time Gait Frequency Step Length Average Average Average
Trials per (s) (Hz) (cm) Accel. Gap - A Accel. Gap - V Accel. Gap - L
Michelle
Walking 0.816 1.23 60 1.91 2.08 1.25
Michelle
Running 0.38 2.63 107 6.72 8.44 7.83
Alexa
Walking 0.69 1.45 61 1.36 1.95 0.66
Alexa
Running 0.47 2.13 107 5.81 9.30 4.41
Tyler
Walking
(additional
trial to prove
model) 0.57 1.75 66 N/A N/A N/A

Page 3 of 10
Personal Data
Personal Height Leg Length Foot to Knee Foot Length
Data (cm) (cm) (cm) (cm) Starting Foot Notes
Patellofemoral
155 90 45 Left Pain
Michelle 24 Syndrome

Alexa 159 98 45 24 Right N/A

Tyler 178 102 51 28 Right N/A

There are 12 different trials that we did extensive analysis on to confirm our hypothesis.
The following graphs show the G-force versus the time for the 12 trials. G-force is actually not a
force, it is an acceleration, and it is the ratio of the normal force over the gravitational force (9.8
m/s^2). Because of this, 1 “G” is equal to 9.8 m/s^2 and the other values of “G” are proportional
to this. For example, when the value is -1 “G” that means that there is a force of 9.8 m/s^2 going
downwards on the accelerometer in the phone, which is just the gravitational force experienced
by all things on Earth. This explains why the G-Force in the y (vertical) direction is -1 when the
person is at rest. In the following trials, the first number denotes the person (1-Michelle, 2-
Alexa). The second number denotes the different trials for each person, with 1-3 being walking
trials and 4-6 being running trials. For example, 1.3 is Michelle walking and 2.6 would be Alexa
running.

Page 4 of 10
One important trend that is visible in these G-force graphs is the range of acceleration in
each direction. As shown by the first table, the range of lateral acceleration is usually the least,

Page 5 of 10
followed by anterior, followed by vertical. This makes sense when walking, because most of the
change in motion (and therefore acceleration) happens in the up and down direction or forwards
and backwards direction. However, beyond this trend within each data set, we could not find a
trend that definitively related height or gait frequency to whether or not these acceleration gaps
would be higher or lower. The shorter person had higher gaps, but since she walked slower but
ran faster, this does not help us prove a correlation with gait frequency. Because of this, we
decided to stick with the walking trials because the speed at which one walks more controlled
than the speed at which one runs/jogs/sprints. We then decided to compare height and gait
frequency instead of the acceleration gap because of the clearer trends.

Predictive Model

Using all this data given, we are able to create some equations (predictive models) that
can predict future data. Ultimately, through the numerous equations we formulated, we relate
height and gait frequency from the data in our trials, because we end up finding the step time,
the inverse of frequency. The equations were created by plotting the data points in Desmos and
creating a line with the points. Below are pictures of the graphs of the equations:

Equation 1 Graph:

Equation 2 Graph:

Equation 1
Let l = step length (cm), Let h = height (cm)

Page 6 of 10
l​ = 0.25​h​ + 21.25
The input in this equation is height, which can be measured easily or is typically already known
by the person, and the output is step length, which can be predicted instead of actually having
to measure a step. This equation means that for every time height increases by 4 cm, step
length increases by 1 cm. This equation’s baseline value was ​h​ = 155 cm, which corresponds to
l​ = 60 cm, and by using this 1:4 ratio, we found that the constant at the end must be 21.25
(y-intercept).

Equation 2
Let t = step time (s), Let l = step length (cm)

t = -0.05l + 3.8
The input in this equation is step length, which can be found in Equation 1, and the output is
step time, which can now be predicted instead of timed. Time can now be used to find gait
frequency by taking the inverse (divide 1 by step time). This equation means that for and
increase in the step length by 20 cm, the step time decreases by 1 sec. The baseline value was
l​ = 60 cm and ​t​ = 0.8 s, which was then used to find the y-intercept of 3.8.

We can either use these equations separately if we want to find each value, or if we want to go
directly to finding the frequency, we can substitute in ​l​ for the second equation, as shown here:

t = -0.05(0.25h+21.25) + 3.8

Gait Frequency = 1/​t


Predictive Model Test:

To ensure that the predictive model works, we then took more data from a third person, Tyler.
We only took basic height measurements and then step length and time, because that is all we
need to test the predictive model. SEE the table at the beginning of the results section for the
exact data. After plugging this new data into Equations 1 and 2, the data verified our equation,
thus helping prove our predictive model. Here is the data in the equation:

Equation 1:
l​ = 0.25(178) + 21.25
l​ = 65.75 cm (Actual Data: 66 cm. Percent Error: 0.4%)

Page 7 of 10
Equation 2:
t​ = -0.05(65.75) + 3.8
t​ = 0.513 sec (Actual Data: 0.57 sec. Percent Error: 10%)

As seen here, both percent errors are relatively small. Equation 1 works very well, and Equation
2 deals with numbers on a much smaller magnitude, thus the higher margin for error.

Discussion/Conclusion

We were able to obtain these results using the app “Physics Toolbox Accelerometer.”
Looking at the data, we found that Michelle had a shorter running step time, but a longer
walking step time compared to Alexa. Looking at the personal data, the longer step time makes
sense for Michelle because she has a shorter leg length compared to Alexa. However, the
running step time for Michelle is much faster because they each had different running speeds.
This discrepancy within the running times is most likely due to the fact that we cannot really
control how fast a person runs. One person might be “jogging” while the other considers it
“running.” Because of this, we decided to focus more on the walking aspect (SEE Results
section above) to answer the discover question. With our equation, we can determine that the
step length will increase by 1 cm when the height is increased by 4 cm. This increase in step
length then correlates to a decrease in step time by about 1 second for every 20 centimeters
more. Please refer to the Predictive Model Section above for the exact equations. Overall, we
see that an increased height related to an increase in gait frequency.
While this model follows our data and is fairly accurate, we always do have some room
for error. As seen in the Results section, our percent error for Equation 1 was only 0.4%.
However, the time percent error in Equation 2 was a bit greater at 10%. This is likely because
the time differences are on a smaller, minute scale, so discrepancies in human recording
produce bigger effects. Additionally, distance (which is what Equation 1 deals with) is a lot
easier to record accurately because there is no reflex time that has to be accounted in. There is
simply a distance and you can record it with a tape measure. Because of this, Equation 1 is
much more precise, but both Equations are valuable in that they can provide insight and show I
general overall trend with certainty.
Overall, we successfully reached our goal for analyzing walking patterns with our two
predictive models. The relationship in the predictive models is also seen in the G-force graphs
(SEE Results). The longer steps correspond to a higher frequency in our equation and we can
see that in the graph, it goes through the cycles (up and down) faster. In the future, additional
tests with more subjects can be done to further back up our model and change it if we find
discrepancies. Additionally, we could do more research into running gaits. We would probably
need to video record running gaits to better analyze them and find some way to standardize the
speed that people are running at, either by choosing people who run the same speed or taking
multiple tests to get one speed that works across multiple people. This project was a great way
for us to learn about analyzing and manipulating data to make equations for trends.

Page 8 of 10
References

“Physics Toolbox Accelerometer.” ​Vieyra Software,​


www.vieyrasoftware.net/physics-toolbox-accelerometer.

“Desmos Graph.” ​Desmos Graphing Calculator​, ​www.desmos.com/calculator​.

Appendix

Here is the raw data from our lab notebook:

Note (correction): the term stride was used to actually denote a “step” - just one foot up and
down, not a step with the right and left feet.

Here is the link to the Google Sheets with the accelerometer raw data and graphs:
https://docs.google.com/spreadsheets/d/e/2PACX-1vRi8UScMVky-e1XKIIjCtNHt2fRhNel_1OnQ
dgbRQMDaOp4FrJbnl0a_EWmtxF3NyN6-80lwcuz24Ng/pubhtml

Page 9 of 10

You might also like