Visualization of Dancer's Emotional State and Cultural Support Using Deep Learning in Breakdance

Abstract

In this research, we developed a system that automatically discriminates and visualizes emotional states of dancers in a natural environment for break dance, one of the dance genres that has become widespread in recent years. We measured acceleration data for each emotional state (valence of emotion), positive and negative, by attaching an inertial sensor to the dancer's body. Then, deep learning using CNN (convolutional neural network) was performed on the acceleration data, and a model for identifying emotional states was built and incorporated into the system. In this research, we introduce the outline of the system and also develop an application that applies the above system and visualizes the relationship between the emotional states of multiple dancers in their battle scenes in real-time . Also, in Hip Hop culture such as break dance, the originality of dancer's movement is very important. Previous studies frequently discussed the relationship between dancer's creativity and their emotional states. Based on these suggestions, we will consider the connection and relationship between the creative process of the dancer's original movement and the emotional state for future cultural support.