LSTM-Based Analysis of a Hip-Hop Movement

Abstract

This paper presents a heuristic approach to analyse the movement of a specific style of hip-hop dance. The data is acquired using a 3-axis IMU and an LSTM network is used to train the data to classify the movement into various phases, each having a distinct sequence of motions, in order to evaluate the performance of the dancer which is otherwise difficult for the human eye. An expert dancer and amateurs' data are compared and the results so obtained can be used to provide constructive criticism to the amateur in order for them to improve.