Motion as a classification signal
This report investigates whether racket kinematics captured during table-tennis swings contain enough signal to classify a combined age and gender label.
Using a modified version of a dataset published on Dryad, I compared three supervised approaches: k-nearest neighbours, support-vector machines, and one-vs-rest logistic regression.
The study examines how racket-motion features contribute to classification and how models with different decision boundaries behave on a real-world, multivariate dataset.