End-to-end Classification of Ballroom Dancing Music Using Machine Learning
by Noemie Voss and Phong Nguyen
Abstract – ‘Ballroom dancing’ is a term used to designate a type of partnered dancing enjoyed both socially and competitively around the world. There are 10 different types of competitive ballroom dancing, each performed to different styles of music. However, there are currently no algorithms to help differentiate and classify pieces of music into their distinct dance types. This makes it difficult for beginner and amateur ballroom dancers to distinguish pieces of music, and know which type of dance corresponds to the music they are listening to. We proposed using an end-to-end machine learning approach to help classify music into different types with efficient and high accuracy. We evaluated four machine learning models and found that a Deep Neural Network with three hidden layers is the model with highest accuracy of 83%. As a result, ballroom dancers will have an easier method of distinguishing between specific types of ballroom dancing music.