Abstract
Basketball is a very popular sport that can benefit greatly from the power of wearable systems and machine learning. It is often said that good form is the most important element in scoring baskets, and wearable systems can help players optimize their shot. By gathering shot accelerometer data, analyzing its features, and using a MATLAB classification learner, a baseline machine learning application can be constructed that can serve as a starting point for more advanced basketball shot analysis systems. A preliminary experiment and analysis conducted on the data collected shows that four different shot forms/types can be classified with an accuracy of 85% using a Quadratic SVM classifier.