Data Collection

In order to analyze the wrist movement for different games, we needed a secure way to attach a sensor to the player's wrist. We first tested with an Apple Watch by using an accelerometer app, but the collected data was very noisy and unreliable.

Data collection setup with Apple Watch

After experimenting with various accelerometers, we discovered that the accelerometer in our phones consistently produced reliable and accurate results. In order to secure the phone onto the player's wrist, we attached the phone onto a glove using velcro and had the player wear it while playing a game.

Data collection setup with phone and glove

We then collected 2 samples per game for the following games,

• League of Legends (MOBA)
• Risk of Rain 2 (TPS)
• Titanfall (FPS)
• Counter Strike (FPS)

giving us a total of 8 datasets.

Motion Model

Before measuring and analyzing wrist movement, we need to determine which wrist motion we are interested in. Based on the image below, most wrist movement produced while gaming will be in the y-direction. Thus, we will be collecting and analyzing the y-acceleration data from the phone's accelerometer.

Shown below is our motion model for y-acceleration of the wrist during different games. Our prediction is that MOBA and Strategy games will result in a significantly smaller max frequency than FPS and TPS games since there is no rapid movement like camara turning and aiming involved. For graphs in the frequency domain, we expect the graphs of FPS and TPS games to peak at a higher frequency.

Signal Processing

Once we have data, we can now analyze it using the Discrete Fourier transform. Discrete Fourier Transform is a common signal analysis method used to extract how much of different frequencies exist in a given signal. Through multiple matrix operations, the signal that was in the time domain is converted to the frequency domain. For the purpose of demonstration, we will be using data from a game of League of Legends.

Y-acceleration data from a game of League of Legends

With the acceleration data, we can perform a fast fourier transform to convert the discrete signal in the time domain to the frequency domain.

Signal in the frequency domain; [-0.5 ≤ x ≤ 0.5]