System capable of automatically classifying the physical activity performed by a human subject is extremely attractive for many applications in the field of healthcare monitoring and in developing advanced human-machine interfaces. Physical Activities can be primitive or composite like sitting-standing-walking-standing-sitting. Its interesting to extract information from physical activities and use this knowledge to calculate metabolic energy expenditure or use in rehabilitation methods. Sometimes, collecting contextual information along with motion can make human-machine interaction more complex and richer in terms of knowledge.

OPTIMISING EXCITEMENT ON ROLLER COASTER A few weeks back I visited Six flags over Georgia. Rather than just having fun like a reasonable human being, I decided to spend time gathering accelerometer data because of my recent interest in activity recognition. I was interested in calculating the G-force and determining which ride or combination of rides to take in limited amount of time to have maximum fun. This can be done by comparing their paths with respect to each other.

I ran all of the experiments presented here using my LG Nexus 4 phone, using the free Physics Toolbox Accelerometer app. This app just measures the x, y, and z accelerations detected by the phone’s accelerometer and records them to a txt file. My phone’s accelerometer also saturates at approximately 4g, so I have no measurements above that value. I kept the phone in my pocket during the rides, which kept it tightly attached to my body, so the accelerations it measured should be representative of what my entire body was undergoing.

I collected data of the following rides – Dare Devil, Goliath, Acrophobia, Ninja, Scorcher, Superman.  I sat on Goliath twice, which is great as it can be used to find ride with maximum similarity among all rides, as experience for learning machine learning algorithm – dynamic time warping.

A balance of Kinetic Energy and Potential Energy is maintained during roller coaster ride. If the tracks slope down, gravity pulls the front of the car toward the ground, so it accelerates. in this case, Potential Energy decreases and Kinetic Energy increases. If the tracks tilt up, gravity applies a downward force on the back of the coaster, so it decelerates. In this, Potential Energy increases and Kinetic Energy decreases. [ ]. Your body can’t feel velocity at all; it can only feel change in velocity (acceleration).
This fluctuation in acceleration is what makes roller coasters so much fun. It constantly changes its acceleration and its position to the ground, making the forces of gravity and acceleration interact in many interesting ways. When you plummet down a steep hill, gravity pulls you down while the acceleration force seems to be pulling you up. at a certain rate of acceleration, these opposite forces balance each other out, making you feel a sensation of weightlessness — the same sensation a skydiver feels in free fall. This is often called “air time”.

Screen Shot 2016-01-07 at 10.32.34 pm

Screen Shot 2016-01-07 at 10.32.43 pmA roller coaster loop-the-loop is a sort of centrifuge. Inertia not only produces an exciting acce­leration force, but it also keeps you in the seat when you’re upside down. As one moves around the loop, the net force acting on your body is constantly changing. It crams so much into such a short length of track. The varying forces put your body through the whole range of sensations in a matter of seconds.


The recording of data started before I sat on the ride, thus it may include noisy data where I am putting phone in pocket, walking to ride, sitting on seat, fidgeting while tying seat belt etc.



I did not take care the orientation of phone while logging data. I have adjusted data so that phone is lying horizontally, facing downwards while I am sitting. That is Z ~ -1, X ~0, Y~0.


The acceleration sensor measures the acceleration along three axes. This 3D vector is handy when one wants to measure orientation of e.g. the gravity acceleration vector. When identifying movements, it is more useful to work with the absolute value of the acceleration because the device may change its orientation during the movement. Therefore we calculate ampl=sqrt( x^2 + y^2 + z^2 )

where x,y,z are elements of the measured acceleration vector along the three axes. Diagrams henceforth will show this ampl value.

Ride length, No. of  Air-Time moments and their length, No. of loops


1. Superman : 


Screen Shot 2016-01-08 at 1.19.23 pm

Screen Shot 2016-01-08 at 1.22.41 pm


2. Acrophobia : 



3. Goliath : 

Ride 1:

goliath ride 1


Ride 2:

goliath ride 2

4. Scorcher: 


5. NINJA : 


6.  Daredevil Daredevil


RESULT : Superman was found to be most exciting ride based on frequency of G force crossing threshold of 2 in the duration of the ride. This was a great fun project  to get the feel of the accelerometer data which laid basis of my interest in activity recognition, time series analysis.

Comments are closed.