Gait Analysis Report
Gait Analysis Report - By Jenna Nottingham, Bailey Bernales, Mia Davis, and Olivia Rigali
Goal:
The goal of this project was to discover a relationship between an individual's gait and their height. Using that information, we then needed to make a predictive model that can estimate a subject’s unknown height when given their gait data. We collected our first set of data for seven individuals using an accelerometer. The data we got from that is the most accurate and not effective for making a prediction model. However, when putting the data in graph form, we were able to identify how specific gait characteristics can impact the results the accelerometer will show. For example, females have sharper hip movements than males, which tends to give them a bigger range of gFx and gFy. Our next attempt to make a prediction model was by using test subjects data from step count to reach 15ft and step length. Then dividing step length by step count to reach 15ft. By making the data all in similar ratios we were able to create a moderately accurate predictive model to find an unknown subject’s height. We then worked to try and create equations to calculate an unknown subject's height. The equations worked extremely well for our seven test subjects collected data. However these formulas were designed to be accurate for those seven test subjects and therefore might not be as accurate for unknown subjects. Overall, I put a lot of effort into this project and definitely will take away a lot of knowledge and confidence knowing I never gave up even when things didn't go as planned for the first few attempts.
Background:
Have you ever thought about the way you walk or run or notice that people around you have distinct ways of walking? As a team we have all individually thought about the way we move as humans. Specific individuals might exhibit unique traits while walking compared to the average human stride. Some people swing their arms more than others, some bounce up and down on their tiptoes, others have interesting postures with either slouched backs or their chest pushed obnoxiously forward. Many walk with their feet pointed inward forming a triangle or outward forming a V. So what can learn from an individual's gait? Gait is defined as a person’s manner of walking. In other words, the pattern of limb movements made during locomotion. Gaits can be natural or formed through specialized training. We know that people of different heights, leg lengths, strides, speed of walking, etc., will all have slightly or drastically different gaits.
Why is gait analysis important? Gait analysis is the systematic study of human motion, using both visual observations and movement measurements, which can have many positive benefits. It can help those with walking abnormalities, such as arthritis, birth defects, bone fractures. In athletics, gait analysis helps athletes learn to be able to move more efficiently for their specific sport, as well as pinpoint specific areas where body issues originate. Through biomechanics, engineers are able to hypothesize, experiment, analyze, and make new discoveries from how people walk.
We collected data of eight different human subjects ranging in heights, both male and female. We recorded subjects height, step count, step length, and seconds worth of data (time). We used an accelerometer to track people’s exact gait movements. This provided us with XYZ points, which represent direction and position of the phone at which acceleration occurred.
Our mission was to discover the relationship between the height and gait frequency for walking humans. As a team, we worked to answer the question: Can you predict someone's height based on their gait? We did this by creating a formula that can calculate an unknown subject’s height, based solely on the information and data (listed in previous paragraph) collected from a subject’s gait. We also constructed a model, representing gaits of different individuals through a 2D and 3D graph.
In each data trail collected we needed to take into account: whether the subject was walking or running, how the phone was held, starting foot, symmetry, variability, dynamicity, and physical characteristics.
Hypothesis:
If you are given an individual's gait data from an accelerator, you will be able to predict approximately how tall they are.
If you are given an individual’s step count to 15ft and step length you will be able to predict approximately that individual's height.
Materials:
8 testing subjects (4 males, 4 females)
20 feet of a straight path
Measuring tape
Accelerometer App
Calculator
Paper/Pen/Chromebook
Duck Tape (optional)
General Procedure:
Record each testing subject’s height.
Have person #1 attach the accelerometer to their upper right thigh. This can be done using tape or be hand held.
Have one person, who is not person #1, be prepared to count steps and another person be ready to measure the stride of the person walking. A third person should be used to record the individual walking.
Have person #1 start at the 0 ft mark. Start the accelerometer and then walk 20ft. It is important to walk in a straight line. Do not run, jump, zigzag, etc.
Once the 20 ft is completed, person #1 turn off the accelerometer and save the data as a .CSV.
Repeat each process until all 8 test subjects have gone.
Calculating the Length/Height Ratio Procedure:
With the data collected, divide the step length (in ft, decimals) by the height of the person (in ft, decimals) to find the step length/height ratio (SL/H Ratio).
Add all test subject's SL/H Ratios together, then divide by the total number of test subjects to find the average SL/H ratio for all of the walkers.
With this ratio, you can then calculate a person’s height by dividing the step length (in ft in decimals) by the average step length/height ratio.
Your answer should be approximately around the person's height.
Data for Experiment 1:
These graphs showed us a few things about how one’s walking traits can impact their gait. First, females tend to have a more noticeable movement of the hips which results in high spikes for gFy, as the phone is being pushed back and forth in a bigger range of motion. If a subject, male or female has a jump in their walk, their change in gFz will be noticeably large and sharp. Lastly, males tend to walk straight on which results in the minimal gFx change compared to the females.
Also females tend to be lower when looking at the y axis. It is hard to tell without data, but it could be possible to determine one’s height based on how low or high their data ranges. For example the lower (more negative) they average, could mean they are around 4.8-5.3 range, and the more positive it gets the taller they are.
Data for Experiment #2
Formulas:
(Step Length ➗ Height) = Step Length/Height Ratio
(Step Length ➗ Average Ratio) = approx. Height
Example:
Jenna’s Height: 5’7 ½” (approx. 5.6)
Step # to reach 20 ft: 7 steps
Step Length: 2.85 ft
2.85 ➗ 5.6 = .50 (Ratio)
Average Ratio: 0.49875
2.85 ➗ 0.49875 = 5.7
5.7 ft is in close proximity to 5.6 ft. If given a human’s SL/H Ratio, we then could figure out the subject's height. However, this is not able to tell us of an unknown subject’s height, since you need the height to get the SL/H Ratio. That would create an endless equation trying to solve for height and SL/H Ratio. This tactic of dividing two individual components of a gait would lead us to our next attempt at discovering how to predict an unknown subject’s height based on their gait.
Data for Experiment #3
Formulas:
(Step Length ➗ Step # to reach 15ft) = SL/S#
How SL/S# can be used for predicting unknown subject heights:
Note: SL/S# is short for “step length divided by step count to reach 15ft”.
We found that the SL/S# would increase with the increase in height from our test subjects. Olivia was our shortest test subject, at 5.33ft. Her ratio was the lowest at 0.2879. Jenna was one of our two middle heighted test subjects, at 5.58ft. Her ratio was 0.3141. Bryce was our tallest test subject, at 6.16ft. His ratio was the largest at 0.3967. Using this information, we were able to begin working on constructing a predictive model.
Using this model one can take a rough estimate of an unknown subject's height when knowing their step length divided by their step count to reach 15ft.
Ex: an unknown subject has a step count to reach 15ft of 7.2115 and a step length of 2.08ft.
Divide 2.08/7.2115 to get an SL/S# of 0.2884
Use the prediction model to see what box it falls in.
Since it is 0.2884 the height is likely to be around 5.35ft
Looking at the data from experiment three, we wanted to challenge ourselves and see if there was another way to predict data using algebraic equations. We came up with a system of equations listed below that can be used to predict a test subject’s height, when given their step count to go 15ft and step length.
Formula to Calculate an Individual's Height:
Height ≈ RSC + SL - C#
RSC = rounded step count to 15ft
SL = individuals step length
C# = Corresponding number
For individuals with a ‘step count to 15ft’ above 7 steps (ex: 7.19 steps)
Round their step count to ‘7’ and plug it in for RSC
Plug in 3.68 for C#
Lastly, plug in that individual’s step length, then solve!
Let's take a look at an example using Olivia’s data.
Olivia has a ‘step count to 15ft’ above 7 steps, so...
plug in 7 for RSC...
and plug in 3.68 for C#
Lastly we will plug in her ‘step length’ of 2.07 for SL then solve!
Height ≈ 7 + 2.07 - 3.68 ≈ 5.39
Height ≈ 5.39ft
If we look back at Olivia’s height in Experiment Two’s data table, Olivia is 5.33. Our estimate is only about 0.06ft off, which can still give us a clear estimate of an individual's height.
For individuals with a ‘step count to 15ft’ above 6.5 but less than 7 steps (ex: 6.58 steps or 6.95 steps)
Round their step count to ‘7’ and plug it in for RSC
Plug in 3.52 for C#
Lastly, plug in that individual’s step length, then solve!
For individuals with a ‘step count to 15ft’ above below 6.5 steps (ex: 6.16 steps)
Round their step count to ‘6’ and plug it in for RSC
Plug in 2.28 for C#
Lastly, plug in that individual’s step length, then solve!
Note: All these formulas were created to generate close to heights of our ten collected subjects. An individual using these formulas on a test subject outside of our ten test subjects might get a less accurate answer. The standard error for our ten data subjects while using these formulas was 0.10ft. However, on other test subjects it may range from .20-.30ft.
Reflection:
After completing the lab, we found that predicting height without data can be a challenging task. Our data from the acceleration app had no correlation or pattern relating to height, which made it extremely difficult to determine a relationship. This could be for a variety of reasons. One, when people are told to walk or know that people are watching them walk, they tend to think too much about walking, and this can mess with their natural gait. They could be forcing short or longer steps, turning their feet in a different direction, pressing off of the ground differently, etc. All of this could affect what data is being collected, and then affect the accuracy of that data. Another reason that data might not be completely accurate is because of the location of where the phone sat on the leg. Some people set the phone right in the belt of their pants and then some people placed it in their pocket, which is further down on the leg. As a team we should have had everyone we used to collect data from tape the phone to a specific location on their leg so the data was coming from a specific spot for every individual. This most likely would have made the g-force more accurate and easier to find a relationship with height.
This lab was a tricky one. Our group had little background on accelerometers and how to read one’s gait. This is a broad topic with heavy loads of information. Being provided with no background information made it challenging to construct a plan, attain the knowledge, collect data, and analyze it in a short time period. Especially when looking up how to use these tools, the internet would provide us with multiple answers with different instructions. Some background knowledge for this lab would have been beneficial, however it did allow us to be challenged and test our creativity when unable to get exact results.
If we were to try this lab again, we would want to test ten boys and ten girls of various heights. We want to make sure that each individual taped the phone to the same spot on their leg, so this time our data is 100% accurate. We also would want to test a longer distance, such as 100ft instead of 20ft/15ft. With more steps we might be able to find new and more accurate formulas, as well as be able to test our old formulas. Also we would want to analyze someone moving at a faster speed, such as jogging or sprinting. We could then discover if we would need new formulas for different motions or if old formulas would still remain accurate.
There is a lot we can take away from this lab as a team. Look at our lab. We had to try four different methods to try and discover a formula to find a correlation between height and gait. Even though each method might not be the most accurate, we put in a lot of effort and time into trying our best to discover a more accurate formula. At the end of the day, no matter what grade our report may end up with, we took away valuable lessons on teaching yourself a subject and learned to stay persistent if the first trial does not work, or the second, or the third. This will leave us as stronger, more productive and confident teammates in the future!