Sign In

Communications of the ACM

ACM TechNews

Machine-Learning Algorithm Watches Dance Dance Revolution, Then Creates Dances of Its Own

View as: Print Mobile App Share:
Inputs to the program Dance Dance Convolution.

Dance Dance Convolution is essentially a machine that can choreograph music.

Credit: Technology Review

Researchers at the University of California, San Diego have developed Dance Dance Convolution, a system that takes as an input the raw audio files of pop songs and produces dance routines as an output--essentially a machine that can choreograph music.

The system initially decides when to place steps, which involves identifying a set of timestamps in a song at which to place steps. Once the timestamps for each step have been identified, the system selects a step to take at each instant.

For this study, the researchers focused on two smaller datasets consisting of recordings plus dance charts, one of which contains 90 songs choreographed by a single author, while the other contains 133 songs each with a single dance chart.

The researchers found the experiments establish the feasibility of using machine learning to automatically generate high-quality dance steps.

From Technology Review
View Full Article


Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account