For this presentation, I let Richard to present our findings and conclusions. The day before when finishing our recordings, we faced with many technical problems and as result we did not have a working prototype to present.
To explain what we did research for, Richard began by showing the video of the movements and then answered the following questions
Why did we choose to work with this movement?
” It offered a fully-embodied movement. It included most of the body parts in order to do the movement. Even though when recording we focused on the thighs as we thought they offered more for the nuance of the movement. We wanted to see if the machine can recognise the small details in the difference of the two different movements we recorded: the normal pitch and the sidearm pitch.”
How does it feel to do the movement?
While experiencing the movement, few things come in the light. Body parts like the neck and the torso are also used in the movement itself. For an observer the torso seems like it is only spinning when the leg is being pushed, but as a participant you have to use the abdominal muscles to move your legs to spin and then pitch the ball itself with the arms.
The movement in itself is actually quite hard. In fact, it isn’t uncommon for professional pitchers to have multiple surgeries during their careers. Our pitching wasn’t nearly as hard as theirs, but we still experienced muscle fatigue and had as a result take brakes.
Similarly, the neck muscles are definitely not something the observer notices since the corresponding motion looks like an ordinary head turn. And you barely feel them being used while performing the movement. But the neck muscle pain experienced in the aftermath proves that there’s much more to pitching movements than meets the eye or even feels like initially. Again, it confirms our original thought that pitching involves most of the body.
What did we learn from this project?
Acquiring knowledge how machine learning works, was not the only lesson that we learned in the past three weeks. As one important key in this project was recording and working with data. Few of the recordings were useless since they did not contain quality data. It became clear that the point is not to only record rather than figure out what exactly can contribute to having better predicting machine.
That quality data can be influenced based on where the sensor is being placed. While we had the phones in the pockets trying to record the movement did not contribute to the training, but switching the sensors to our hands gathered better data.
Pink Prototyping 