We used an accelerometer to measure how violent the ball is shaking. However, the position (tilt) of the acceleromoter cannot be pre-determined, thus the influence of gravity cannot be offset. We decided to record 10 value of acceleration along the 3 axis and compute the variance to determine the "shakiness".
We experimented how "Shakiness", or "violence", can be derived from the variation of acceleration along 3 axis: Maximum variance, Average variance, or Minimum variance. It turned out that Minimum variance was the most effective indicator of shaking.
Is this a good/useful/informative piece of content to include in the project? Have your say!