Poi Style Classifier

Project created by: Alex George, Mo Ampane, and Belén Hutchins

Our Purpose

The goal of this project is to identify the characteristic frequencies of different poi styles and use them to identify distinct characteristic frequencies between “flowy” and “stally” styles of spinning. For the purposes of this project, flowy is defined as the more continuous movement of both the poi and the spinner, and stally is defined as sharper, broken-up movements with the poi.

What Are Poi?

Poi is a performance art style originating from the Maori people of New Zealand. Poi is also the name of the prop used as the word literally means ball on string. There are a variety of different styles of movement and tricks that performers learn and develop over time. Common ones and the ones used in this project are stalls (sudden stop of movement before changing direction, either parallel or perpendicular to the ground), fountains (a trick that continuously moves the poi across and over the body), corkscrews (a trick that continuously rotates the poi while moving up and down), and windmills (starting spinning the poi on a slight delay from each other, bring them over your shoulders while rotating your torso, then bringing the poi back down).

Flowy Snippet.mov

Flowy Spin Style

Stally Snippet.mov

Stally Spin Style

Value Creation

It is no surprise to any Oliner that people who spin fire are always looking for new things to learn and ways to improve. By tracking how a spinner moves while practicing, they can learn more about their spinning style and how consistent they are. This way, any spinner is able to better understand what is needed to achieve different goals. For example, if a spinner wants to be more consistent with their style and movement, they can check to see if the data collected reflects that. In addition, this data can help sort people for group preferences. If everyone performing has the same or similar spinning style, the performance will look more cohesive.

How It Works

Use our poi to determine your spin style
Spin as you normally would, or try to aim for a specific style
Based on the acceleration data that our poi collect, our algorithm will tell you if your spin style was more flowy (continuous motion) or stally (more breaks with abrupt changes in movement)
Use that information to choose who you peform with, change or focus your spin style, or whatever else you can think of!

Proof of Concept & Next Steps

The classifier our algorithm runs on was trained on ten sets of features, five flowy and five stally. When tested on two sets of external features it classified both correctly, the first as flowy and the second as stally. In the future, we would like to gather more testing data to further test the limits of our classifier and get a better idea of how accurate it is, but based on its current accuracy we are comfortable saying that there is a solid proof of concept.

Further next steps of this product would be to test it with a wider user base. People spin in many different ways and with different levels of skill. It is possible that we would have to break this analysis further down into more styles.

We would also want to gather a wider variety of testing data. Currently, our testing data focuses primarily on “stally” tricks or “flowy” tricks, and not the overall integration of them, which is typical during a performance. While there is some overlap in our testing data (transitioning between “stally” tricks sometimes was done through more “flowy” tricks) further testing, data collection, and labeling needs to be done to determine where the threshold between these two styles lie.

A useful feature of this product would be to recommend tricks for a user to learn that match their style. Finally, data collection would be expedited by a mockup of an actual product - a poi head with a built in accelerometer.

Page updated

Google Sites

Report abuse